Railroad Bailouts in the Great Depression

Abstract

The Reconstruction Finance Corporation and Public Works Administration loaned 46 railroads over $802 million between 1932 and 1939. The government’s goal was to decrease the likelihood of bond defaults and increase employment. Bailed-out railroads did not increase profitability or employment. Instead, they reduced leverage. Bailing out a railroad had little effect on its stock price, but it resulted in an increase in its bond prices and reduced the likelihood of a ratings downgrade. However, bailouts did not help railroads avoid defaulting on their debt. We find some evidence that manufacturing firms located close to railroads benefited from the bailouts.

Full text

1
Railroad Bailouts in the Great Depression*
LYNDON MOORE
Monash University
GERTJAN VERDICKT
KU Leuven
Abstract
The Reconstruction Finance Corporation and Public Works Administration loaned 46 railroads over
$802 million between 1932 and 1939. The government’s goal was to decrease the likelihood of bond
defaults and increase employment. Bailed-out railroads did not increase profitability or employment.
Instead, they reduced leverage. Bailing out a railroad had little effect on its stock price, but it resulted
in an increase in its bond prices and reduced the likelihood of a ratings downgrade. However, bailouts
did not help railroads avoid defaulting on their debt. We find some evidence that manufacturing firms
located close to railroads benefited from the bailouts.
KEYWORDS: New Deal, Reconstruction Finance Corporation, Public Works Administration,
Railroads
JEL CLASSIFICATION: H81, L92, N22, N42
* We thank Nabil Bouamara, Fabio Braggion, James Brugler, Viet Cao, Julio Crego, Abe de Jong, Toby Daglish, Hans
Degryse, Rik Frehen, Carola Frydman, Neal Galpin, Will Goetzmann, Florian Hoffmann, Peter Koudijs, Joseph Mason,
Chris Meissner, Magdalena Rola-Janicka, Yulong Sun, Stijn van Nieuwerburgh, Marno Verbeek, and Barry Williams for
helpful comments. We are also grateful to individuals at the Australian National University, Financial History Workshop
(Antwerp), Irish Academy of Finance, KU Leuven, Latrobe University, LSE-UC Davis Economic History Seminar,
Massey University, Monash University, Oslo Metropolitan University, Paris School of Economics, Tilburg University,
Queen’s University Belfast, University of Antwerp, University College Dublin, and University of Melbourne for helpful
suggestions. Hannah Merki and Enzo Peeters provided excellent research assistance. We thank Gustavo S. Cortes for
sharing his data.
Lyndon Moore (corresponding author), lyndon.moore@monash.edu
Gertjan Verdickt, gertjan.verdickt@kuleuven.be

2
1. Introduction
The Reconstruction Finance Corporation (RFC) was created by President Hoover in early 1932 during
the depths of the Great Depression. The objective of the RFC was to “make temporary advances upon
proper securities to established industries, railways and financial institutions which cannot otherwise
secure credit, and where such advances will protect the credit structure and stimulate employment.”
1
The Corporation approved $3.9 billion in loans from 1932 until 1939. We call these loans ‘bailouts’
because they were provided at below-market interest rates and were intended for companies that could
not access credit from commercial sources.
2
Most of the RFC’s loans went to the financial sector (see
e.g., Mason (2001), Calomiris, Mason, Weidenmier, and Bobroff (2013), and Butkiewicz (1995,
1997)), but 8.6% ($802 million, which included rolled-over loans) was approved for the nation’s
railroads during the eight-year period. We explore whether the RFC’s support, along with more
limited assistance from the Public Works Administration (PWA), resulted in the achievement of the
RFC’s twin objectives for the railroad sector. First, did bailouts protect the “credit structure” of the
railways, meaning did a government loan help a railroad to avoid default? And second, did bailouts
help railroads to increase employment?
We find no evidence that bailouts were successful in reducing railroad defaults, although they reduced
the likelihood that the railroads’ bonds would receive a ratings downgrade, and they permitted
railroads to reduce their leverage. We also find no evidence that government loans were successful
in improving bailed-out railroads’ employment. However, when a newspaper reported that a railroad
applied for a government loan, its bond prices jumped by 0.9% that day. And in the nine days
surrounding the application announcement, railroads’ bond prices experienced an abnormal return of
4.2%. Similarly, news of a loan approval coincided with a 0.3% increase in bond prices on the day of
the approval and a 1.6% abnormal return in the nine days surrounding the announcement.
Government loan applications and approvals are not robustly associated with abnormal returns for
the railroads’ equity.
Although New Deal railroad assistance was not explicitly aimed at the railroads’ customers, we find
evidence that firms located in the same city or town as a bailed-out railroad benefited from news of a
forthcoming railroad loan. Manufacturing firms that had significant operations overlap with the
1
RFC Final Report (1959), page 1.
2
In 1932 and 1933, RFC loans were extended at the same interest rate as Federal Reserve loans to member banks. See
https://www.federalreservehistory.org/essays/reconstruction-finance-corporation (accessed April 26, 2022).

3
assisted railroad experienced a 0.9% abnormal return upon announcement of the bailout, compared
to a 0.4% abnormal return for manufacturers with low levels of overlap with that railroad.
An advantage in studying railroads during the Great Depression is that most railroads—unlike
banks—had publicly traded debt and equity. We can, therefore, study the immediate impact of
government assistance on security prices. In addition, there is extremely granular data on railroads,
which allows us to know where the railroads operated, the products they transported, and their
employment levels. We also have annual financial statements and monthly revenue reports on the
railroad firms that received government loans, which are some of the largest firms in existence. The
Baltimore and Ohio and the New York Central railroads, for example, had balance sheets in excess
of one billion dollars and operated more than 5,000 miles of track. Furthermore, details of government
railroad loans were quickly made public by the railroad regulator, the Interstate Commerce
Commission (ICC), and reported in the media. It was, by contrast, impossible to observe the
immediate impact of government loans in most sectors during the Great Depression. Loans to banks,
farms, and industrial firms were largely kept secret, and financial claims on these firms were not
usually traded in liquid financial markets.
Railroad bailouts were not intended for railroads in the most precarious financial positions since the
RFC was obliged to ensure that loans were “adequately secured.” Only nine applications were
rejected. Railroads that were successful in obtaining a government loan likely differ from railroads
that did not receive government aid. Although we condition our results on the publicly available
characteristics of railroads, it is likely that railroads also differed along unobservable dimensions. To
address this issue, we take advantage of the political process that was inherent in RFC decision-
making. RFC directors were appointed by the President and confirmed by the U.S. Senate. Political
considerations appear to have been important in the decision-making process, as bailouts were more
likely to be granted to railroads that operated in the home states of RFC directors. When we use the
composition of the RFC board as an instrument for RFC loans, we still find no beneficial effect of
loans on railroad employment, profitability, or debt repayment.
Policymakers are often willing to provide aid to the banking system during a financial crisis (see e.g.,
Bordo and Schwartz (2000), Grossman and Woll (2014), and Lucas (2019)). The objective of such
aid is to prevent a reduction in bank loans to the real economy and a resultant recession. Former U.S.
Treasury Secretary Hank Paulson looked back in 2018 on the Global Financial Crisis and said, “I
would look into the abyss and just see food lines, see a second Great Depression, wondering if one

4
more institution went down how would we put it all back together again.”
3
Kelly, Lustig, and Van
Nieuwerburgh (2016) show that options markets (correctly) anticipated government assistance to the
financial sector, but not other sectors, during the Global Financial Crisis. Direct government aid to
the real economy has been rarely attempted during a crisis, although direct aid was a big part of many
governments’ COVID-19 responses (see e.g., Cirera et al. (2021), Elenev, Landvoigt, and Van
Nieuwerburgh (2021), and Granja, Makridis, Yannelis, and Zwick (2020)). A crisis in the financial
sector, however, can easily spill over into the real sector, as Benmelech, Meisenzahl, and Ramcharan
(2017) demonstrate in the market for automobiles and automobile loans during the 2007-2008 crisis.
The effectiveness of financial system bailouts has been studied extensively, both theoretically and
empirically (see e.g., Acharya, Drechsler, and Schnabl (2014), Aghion, Bolton, and Fries (1999),
Berger and Roman (2015), Berger, Bouwman, Kick, and Schaeck (2016), Diamond and Rajan (2004),
Diamond and Dybvig (1983), Duchin and Sosyura (2012), Ennis and Keister (2009), and Gorton and
Huang (2004)). Problems in the financial system during the Great Depression have received much
attention (see, among others, Friedman and Schwartz (1963), Bernanke (1983), Calomiris and Mason
(1997, 2003), and Benmelech, Frydman, and Papanikolaou (2019)). The consensus is that conditions
in the financial and banking sector worsened the real effects of the Depression.
Government aid to non-financial firms during a crisis has received almost no attention in the academic
literature. Faccio, Masulis, and McConnell (2006) find that politically connected firms were more
likely to be bailed out around the time of the Asian financial crisis, especially if the national
government had received an IMF or World Bank aid package. The authors conclude that bailed-out
firms that were politically connected continued to underperform non-bailed-out firms in the same
industry following the bailout, as measured by the return on assets (ROA). However, the ROA for
non-connected firms improved relative to same-industry peers after a bailout. The study does not,
therefore, fully determine whether real-sector bailouts are good public policy in a crisis.
Goolsbee and Kruger (2015) argue that the bailouts of General Motors and Chrysler in 2008 helped
to reduce the economic downturn in the U.S. They conclude, “The rescue has been more successful
than almost anyone predicted at the time.” Their study is necessarily restricted to two firms since the
remaining Troubled Asset Relief Program (TARP) funds went to the financial sector. Berger and
Roman (2017) investigate economic spillovers following TARP bailouts of U.S. banks. They find
3
https://www.cnbc.com/2018/09/12/bernanke-paulson-and-geithner-say-they-bailed-out-wall-street-to-help-main-
street.html

5
that in the counties in which banks received more TARP funds, there was better net job creation—
perhaps because TARP recipients passed on more generous loan terms to their customers (see Berger,
Tanakorn, and Roman (2019)). Assistance first went to Wall Street before going to Main Street. While
Berger and Roman study indirect assistance to the real sector we examine direct loans (at preferential
interest rates) from the government to industry. In contrast, Faccio, Masulis, and McConnell (2006)
study direct bailouts from the government to firms.
It is important to study assistance to non-financial firms in a crisis since there are important
differences between financial firms and non-financial firms. Financial firms, for example, can
experience ‘runs’ on the demand deposits that support a bank’s assets. In addition, a financial firm
can dramatically change its business operations by reducing loans (to preserve cash reserves) or
taking on increasingly risky loans to ‘gamble on resurrection’ (see e.g., Hellmann, Murdock, and
Stiglitz (2000) and Dewatripont and Tirole (2012)). In contrast, non-financial firms face few of these
issues. U.S. railroads, for instance, often issued 50-year bonds to finance their operations, so there
could be no ‘run’ on the railroad’s debt unless their bonds were close to maturity (see Benmelech,
Frydman, and Papanikolaou (2019)). Furthermore, taking on increased risk in a crisis is difficult for
railroads (or non-financial firms in general) since tracks, and other real assets, are fixed and costly to
divert in the search for new customers.
We discuss the economic environment that led to the creation of the RFC and the Corporation’s
structure in section 2. We describe our data and sources in section 3. We present our main results in
section 4 with robustness checks in section 5. We conclude in section 6.
2. The Great Depression and the Reconstruction Finance Corporation
The Great Depression was an unprecedented period of economic and financial collapse worldwide.
It struck the U.S. particularly severely, with peak to trough industrial output falling 40% by late 1931
and GDP still 25% below trend six years after the recovery began (see Cole and Ohanian (2004) and
Ohanian (2009)). There were several waves of banking crises in the early 1930s (see Bernanke (1983)
and Friedman and Schwarz (1963)). In response to the weak economy and runs on troubled banks,
President Hoover reluctantly created the Reconstruction Finance Corporation in January 1932, which
was a component of what came to be known as the ‘New Deal.’ The RFC was initially permitted to
loan to financial firms and railroads; loans were later permitted to farms, state and local government,
infrastructure projects, and industrial loans. The RFC’s initial capital stock came from a $500 million
appropriation from the Treasury. While it obtained the bulk of its additional funding by issuing notes

6
to the Secretary of the Treasury, a very small part of its operations was provided for by direct
borrowing from the public.
The New York Times reported on December 19, 1931, that Hoover believed that “the plight of the
American railroads is only temporary and that they will be able to work themselves out of the
depression.”
4
The United States had experienced severe railroad defaults during crises in 1873, 1884,
and 1893, in which multiple large lines defaulted, resulting in significant drops in railroad
employment (see Schiffman (2003), Giesecke, Longstaff, Schaefer, and Strebulaev (2011), and Cotter
(2021)).
Part of the rationale for providing aid to railroads was that many railroads were not capable of
repaying their maturing bonds, and it would be exorbitantly expensive for them to obtain alternative
funding from the banking sector. In late 1931, Daniel Willard testified in the Senate that railroads
“cannot borrow money from banks at less than 8 or 9 per cent interest” when most maturing bonds
had coupon rates of around 4 percent.
5
Figure 1 shows the number of new railroad bonds issued by
year. Treasury Secretary Andrew Mellon saw the role of the RFC as to provide “a stimulating
influence on the resumption of the normal flow of credit into channels of trade and commerce.”
6
The
Reconstruction Finance Corporation Act became law on January 22, 1932. The initial board of
directors was appointed on February 2, 1932, and the first applications were received on February 5,
1932.
RFC loans to railroads were limited to three years duration, had to be ‘adequately secured’ by
collateral, and were restricted to railroads that could not obtain funds on reasonable terms—although
no definition of “reasonable terms” was provided. Many railroad loans were made for a three-year
duration, and 83.5% of loans in our sample were rolled over. Railroads that were reorganizing under
bankruptcy protection were also eligible for RFC loans. Since railroads’ normal operations were
regulated by the ICC, both agencies had to approve government loans to railroads.
Over the entire period of the RFC’s existence (1932-1957), the agency recovered 97.99% of the
nominal value of the loans (see RFC Final Report, p. 163). We halt our examination of RFC loans in
1939, since the Great Depression is usually considered to have been over by the end of the 1930s.
4
New York Times, December 19, 1931, page 4.
5
New York Times, December 23, 1931, page 16.
6
New York Times, December 24, 1931, page 6.

7
Although disclosure of RFC loans to banks was sporadic, the ICC had a policy of full and timely
disclosure of railroad loans. All railroad loans were publicly disclosed at or near the time of loan
application and approval. Loan information was sometimes, however, delayed slightly. The
Baltimore and Ohio Railroad’s loan application, for example, was kept secret for 10 days in August
1932. In addition, railroads appear to have been occasionally permitted to quietly drop a loan
application without being formally rejected. We show the distribution of RFC railroad loans over
time in Figure 2. We depict the geographical distribution of loans by state in Figure 3.
The composition of the RFC board was determined by the President and confirmed by the Senate.
The initial board’s ex-officio members were the Secretary of the Treasury, the Chairman of the
Federal Reserve Board of Governors, and the Farm Loan Commissioner. Directorships were balanced
by party affiliation, and care was taken for the directors to come from different regions of the U.S.
We read press reports and online biographies of the RFC directors to assign, where possible, the
directors’ ‘home’ states. For example, the New York Times reported that two members of the initial
RFC board would be “two Democrats from the Southwest, Harvey C. Couch of Arkansas and Jesse
H. Jones of Texas.”
7
The RFC final report also describes the home state of most of its directors. We
find that the home states reported in the newspaper align with the RFC’s designations. We document
the composition of the RFC board in Table 1, panel A. Most of the appointed RFC directors were
businessmen, and four were former U.S. senators.
New Deal funding decisions are generally considered to have been at least partly politically motivated
(see e.g., Wright (1974), Wallis (1987), and Fishback, Kantor, and Wallis (2003)). The RFC’s
decisions were similarly criticized. In April 1932, Representative La Guardia claimed, “Everybody
in the country knows a private wire from J. P. Morgan to the headquarters in Washington dictates the
[RFC’s] policy.”
8
The RFC’s initial president, former Vice President Charles G. Dawes, was heavily
criticized by Senator Brookhart of Iowa for having loaned over $80 million to Dawes’ own Chicago
bank.
9
In our analysis, we demonstrate that RFC railroad loans were also partly determined by the
geographical origins of the RFC board. We use the composition of the RFC board at the time loans
were made as an instrument for loans.
The Public Works Administration also made government loans to railroads from late 1933 until early
1936. PWA loans tended to be smaller than the RFC’s disbursements, and they were often used for
7
New York Times, January 26, 1932, page 1.
8
New York Times, April 7, 1932, page 2.
9
See New York Times, September 16, 1932, page 2.

8
capital expenditure rather than to service the railroad’s debt. Since money is fungible, however, we
consider both RFC and PWA loans in our analysis. PWA loans only comprise around 10% of our
sample by value, and 15% of our sample by number.
3. Data
3.1 Data sources
We study the Class I railroads of the continental United States. Class I railroads owned over 90% of
the nation’s tracks by length, they employed roughly 98% of railroad employees (representing 3.4%
of the United States’ labor force), and they carried over 99% of the revenue-ton-miles of all U.S.
railroads in 1929. We collect balance sheet, profit and loss, track network, and employment data for
these railroads from the annual reports of the Interstate Commerce Commission, Statistics of Railways
in the United States.
We compile annual statistics for each railroad’s freight revenue sources (i.e., agricultural, animal,
mining, forestry, merchandise, or manufacturing items) and monthly revenue from Moody’s Manual
of Investments – Railroad Securities. Data on freight revenue is important because some railroads
concentrated on transporting a narrow range of products. For example, the Monongahela Railway
Company, the Montour Railroad Company, the Pittsburgh and Shawmut Railroad Company, and the
Bessemer and Lake Erie Railroad Company obtained over 90% of their freight revenue from mineral
products. As a result of the railroads’ varied exposure to product markets, the Great Depression
affected railroads unevenly. We use the ICC’s classification of railroads’ geographical region (i.e.,
New England, Great Lakes, Central-Eastern, Pocahontas, Southern, Southwestern, Central Western,
and Northwestern).
If a railroad had publicly traded equity, we gather stock prices from the Center for Research in
Security Prices (CRSP). Many railroads did not have publicly traded equity, usually because they
were fully owned by another railway or a related industrial firm. We compile price data on the two
most liquid bonds per railroad.
10
We obtain bond prices from the section, ‘Bond Sales on the New
York Stock Exchange,’ in the New York Times. We classify a railroad as being in default if it failed
to meet a coupon or principal repayment, or if it in any way changed the terms of the issue.
11
Data on
10
There are 46 railroads with listed equity and 72 railroads with bonds.
11
For example, extending the maturity of the bond, reducing the coupon rate, or exchanging the initial bond for another
security.

9
bond ratings, coupons, amounts outstanding, and maturity comes from Moody’s Manual of
Investments – Railroad Securities. We use Moody’s index of daily railroad bond prices, reported in
the Commercial and Financial Chronicle, as a proxy for the market return on railroad bonds.
We link the track network of each railroad with two city-level data sources. First, we hand collect
data on factories operated by NYSE-listed manufacturing firms from Moody’s Manual of Investments
– Industrial Securities. In total, there are 471 manufacturing firms that have data on factories reported
in Moody’s. Second, we obtain city-level building permits data from Cortes and Weidenmeir (2021).
The value of these permits is based on the costs of new commercial and residential buildings for 215
cities across the U.S., taken from Dun & Bradstreet’s Review.
3.2 Bailouts
To identify railroad bailouts, we search the New York Times for the phrases “Reconstruction Finance
Corporation” or “Public Works Administration” from January 1932 until December 1939. We collect
the date of loan applications, approvals, and rejections, as well as the name of the railroad, and the
size of the loan. We define an “approval” as the date on which it became clear that the RFC would
approve a loan. Informal approval could come before an application. For example, the head of the
RFC would occasionally publicly state that the Corporation would be willing to grant a loan to a
certain railroad if it were to apply. On February 16, 1939, the New York Times quoted RFC chairman
Jesse H. Jones: “The RFC was willing to lend $5,000,000 to the Minneapolis and St. Louis Railroad
if its reorganization plan is approved by the courts and the Interstate Commerce Commission.” Not
all approvals, then, were preceded by an application. Similarly, most, but not all, applications are
followed by a newspaper report of an approval or a refusal. The approvals/rejections of small
applications may not have been newsworthy enough to be reported, and a single approval was
occasionally announced for a railroad that had submitted multiple applications in prior months. There
were also several occasions when railroads would revise the size of their loan request while the
application was under consideration. Therefore, our designations of “applications” and “approvals”
occasionally combine multiple information events. Loan decisions were made quickly, usually taking
a couple of weeks to a month or two.
We collect information on the composition of the Reconstruction Finance Corporation board from
the Final Report on the Reconstruction Finance Corporation (1959). We obtain information on the
home state of the board members from reports in the New York Times and the RFC final report. Bank
capital in default comes from the Annual Report of the Comptroller of the Currency. Bank capital per
state comes from Flood (1998).

10
4 Results
4.1 RFC Board composition
In Table 1, panel B, we investigate the composition of the RFC board. The composition was supposed
to be balanced by party affiliation and geographically diverse. However, it is possible that the
appointment of RFC directors was partly determined by economic conditions in the home state of the
director or even by financial conditions in the railway sector in their home states. Larger states were
more likely to have RFC directors, and we find evidence that directors were less likely to be appointed
if there was already a director from the same state. Our results show that the appointment of directors
is not robustly associated with economic conditions in the directors’ home states. Therefore, concerns
are alleviated that causality runs from the poor economic conditions of railroads to the appointment
of RFC directors and thence to more railroad bailouts.
4.2 Summary statistics
In Table 2, we present our summary statistics on railroads. We divide railroads into those that were
“bailed-out”—which we define as having received at least one loan from the RFC or the PWA
between 1932 and 1939—and those that were not bailed-out. In Panel A, we show that there are large
differences between the bail-out recipients and others. Bailed-out railroads had less cash to total assets
(a mean of 1.5% of assets vs. 2.4% for non-bailed-out railroads), were slightly less levered (mean
long-term debt to total assets of 41.6% vs. 43.9%), were less profitable (a mean net income to total
assets of 0.9% of total assets vs. 1.6%), and had less volatile operations (monthly volatility of 14.1%
vs. 21.4%). Bailed-out railroads were much larger (mean total assets of $260.7 million vs. $22.7
million), employed more people (a median of 12,750 vs. 1,100), had a higher wages component of
costs (63.6% of operating expenses vs. 61.6%), and had more (same-state) connections to the RFC
board. On average, bailed-out railroads operated in 1.481 states with an RFC director vs. 0.745 for
non-bailed-out railroads.
Since the statistics in Panel A combine observations before and after the loans, we examine ex-ante
differences between loan recipients in Panel B. We find that the differences between bailed-out
railroads and non-bailed-out railroads in 1929 mirror those in the full sample. Loan recipients had
less cash, higher employment and assets, and were less volatile than railroads that did not receive a
bailout.

11
In Panel C, we split the bailed-out railroads into two groups: those that received a single loan from
the RFC or PWA and those that received multiple bailouts between 1932 and 1939. The railroads that
received multiple bailouts tended to have less cash (mean 1.4% of total assets vs. 1.8%) and much
lower profitability (0.8% of total assets vs. 1.6%). The companies that obtained multiple bailouts
tended to be larger (mean size of $289.6 million vs. $126.3), employ more people (a mean of 13,413
people vs. 4,196 people), focus more on passengers (11.5% vs. 8.5% of total revenue at the mean)
and have more (same-state) connections to the RFC board (on average 1.662 vs 0.736).
4.3 Bailout recipients
We use a two-step model to investigate which railroads received government bailouts. In the first
stage, we run a probit model of a railroad’s application (Application equals one) on time-varying firm
characteristics. We find that railroads with less cash and those that were younger, larger, and less
reliant on passenger revenue carriers were more likely to apply for a bailout. A one percent increase
in the railroad’s age decreases the probability of applying for a bailout by 5.63%.
In the second stage, the dependent variable is an indicator that equals one if a railroad received a
bailout (Approval) at least once in a certain year. This two-step regression allows us to address
potential selection biases that may arise from the endogenous approval of railroad bailouts. Among
railroads that apply for a loan, the number of political connections was a critical factor in determining
who was approved. An additional RFC Connection increased the probability of receiving a bailout
by 21.1% (column 2). This result is robust to using an OLS (column 3) or Heckman (column 4)
specification. The political process behind RFC/PWA loan approvals was vital for a railroad to
receive funding, in contrast to economic factors such as leverage, profitability, or bonds that were
close to maturity.
4.4 Market reactions to bailouts
We examine the reactions of a railroad’s stock and bond prices to news of bailout applications and
approvals. In Table 4, we compute abnormal returns (AR) and cumulative abnormal returns (CAR)
on railroad debt and equity. We choose two benchmark bonds for each railroad, or one if there was
only a single bond issued. We select the most liquid bonds traded on the New York Stock Exchange.
We compute abnormal returns as the return on the stock or the bond less the CRSP market return or

12
Moody’s railroad bond index return, respectively.
12
Mason and Schiffman (2003) calculate that in
1931, 31% of railways’ debt was held by insurance firms, 17% by banks, 4% by foundations and
educational institutions, and 7% by other railroads, with the remainder held by “other” investors.
We find statistically significant abnormal returns of 0.9% for bonds on the day a loan application was
announced and an abnormal return of 0.3% on the day an approval was announced (see Panel A).
There is no statistically significant AR on refusal announcement dates, although there is a -5.9% AR
the next day. In the window around the news release (t-4 to t+4), we find CARs of 4.2% (applications),
1.6% (approvals), and -2.5% (refusals), although the refusal return is statistically insignificant.
Since many railroads applied for (and were granted) multiple loans, we investigate the differences
between the initial loan and subsequent loans. Substantially more private information is likely to have
been conveyed to the market by a firm’s initial revelation that it desired federal government financial
assistance. An application announcement for the first bailout is associated with a 9.8% bond CAR
from t-4 to t+4, although the 2.6% AR on day zero is insignificant (see Panel B). An approval
announcement for the first bailout has a 0.9% bond AR on day zero (with a 4.0% CAR from t-4 to
t+4), all statistically significant. Subsequent bailouts are reflected in more subdued bond responses.
A second or subsequent application has a 0.4% bond AR on announcement day (2.4% from t-4 to
t+4), both statistically significant. A second or subsequent approval has a 0.2% AR (insignificant) on
day zero and a statistically significant 1.1% CAR from t-4 to t+4.
In contrast to the response of bond prices, there is little statistically significant movement of stock
prices in response to bailout news events (see Panels A and B). Most estimates are statistically
insignificant, including the extremely large abnormal return of -23.9% on the day of the initial loan
approval. News of subsequent RFC approvals resulted in a 1.5% AR on the announcement date and
a 2.3% CAR (both statistically significant) from t-4 to t+4.
4.5 Determinants of announcement returns
We examine the association between a railroad’s characteristics at the time of the bailout and its
announcement returns. In Table 5, we regress the CAR of the railroads’ bonds and equity (from t-4
to t+4) on firm and bailout variables. We find few railroad characteristics that are robustly associated
12
We only hand collect bond prices in a narrow window around RFC announcements. Therefore, we are unable to estimate
a market model for railroad bonds. To maintain consistency in our measurement of abnormal returns between bonds and
stocks, we compute abnormal returns for railroad stocks in the same manner. Effectively, we assume that alpha equals
zero and beta equals one in the market model.

13
with security returns. Most characteristics are insignificant and change signs depending on whether
we examine applications versus approvals or bonds versus shares. Railroads with more leverage
experience substantially worse returns on their equity upon announcements of loan applications,
perhaps because a loan application indicated the railroad would struggle to service its debt, and
therefore that equity was next to worthless. A one standard deviation increase in leverage corresponds
to a 9.60% larger CAR. Railroads with more employees tended to have lower announcement returns,
perhaps because market participants expected government pressure on the railroad to maintain
employment.
13
A one standard deviation increase in employment is associated with a 2.03% smaller
CAR. In Panel B, we distinguish between initial and subsequent bailouts. Again, we fail to find robust
relations between characteristics and abnormal returns, although higher leverage was generally
associated with worse returns for debt and equity.
5 Effectiveness of government bailouts
5.1 Bond defaults
We now turn to the central question: did the RFC achieve its stated objectives? We start by examining
if it achieved its first objective, protecting the “credit structure” of the financial system. All else equal,
an RFC loan should have made a railroad less likely to default on its debt. Jones (1951) claims that
RFC funding reduced railroad defaults by half, whereas Schiffman (2003) and Mason and Schiffman
(2004) claim that bailouts at least delayed defaults. However, defaulting on debt is partly a choice,
and Mason and Schiffman argue that bankruptcy “brought relief from high fixed charges that were
often a principal cause of financial distress” (p. 61). In Figure 4, we plot Kaplan-Meier (1958) graphs
with the cumulative probability of failure for bailed-out vs. non-bailed-out railroads. We observe that
railroads that received a bailout are associated with a higher hazard rate of bond defaults and that this
difference increases over time. In Panel B, we show that the higher default rate for bailed-out railroads
survives the inclusion of control variables. The granting of a below-market-rate loan, ceteris paribus,
is a good event. Therefore, higher default rates for a bailed-out railroad suggest that unobservable
factors are likely influencing a railroad’s performance and the government’s proclivity to grant
bailouts.
13
An alternative explanation is that investors may have perceived that railroads that received bailouts would be more
generous with their employees’ compensation.

14
We assess the effects of bailouts on bond defaults in Table 6. In column 1 we run a probit model of
defaults in which the dependent variable equals one if a railroad defaulted on its bonds in that year.
We examine if a government loan Approval in the previous year is associated with the railroad
defaulting upon its debt. We attempt to capture railroad unobservables by including bond rating fixed
effects from Moody’s. In this era, Moody’s only released ratings once per year in its annual investors’
compendium. Most railroads had multiple bonds and bond ratings so we use the rating of the bond
closest to maturity. Overall, we show that lower net income, lower cash to assets, more bonds
maturing in the depths of the Depression (1930 to 1934), and youth are associated with a higher
likelihood of default. Government loan Approval does not have a statistically significant relation with
defaults.
In column 4 we run a related probit model of defaults, but the dependent variable now equals one if
a railroad defaulted on its bond in that year or the following two years. This offers a longer-run
investigation of how a government loan Approval is associated with a railroad’s likelihood of default.
We find that Approval has a significantly positive correlation with the probability of railroad default
(at the 5% confidence level). Indeed, getting an RFC or PWA bailout is associated with a default rate
of 6.39%, all other characteristics at sample means, relative to an unbailed out railroad’s default rate
of 1.51%. This finding is in line with the Kaplan and Meier graphs in terms of magnitude.
14
However,
railroad bailouts are unlikely to be awarded at random, and a selection effect is likely to be present.
Hence, we turn to an instrumental variable approach to determine if bailouts have a causal effect on
railroad defaults.
5.2 Instrumenting for bailouts
Our major concern in determining if bailouts aided railroads is that there are likely to be omitted
variables in our econometric specification that partly affect a railroad’s financial performance.
Railroad management and policymakers on the RFC board were likely to have had access to
information that we do not. For example, a railroad that had tried but failed to obtain bank or Wall
Street assistance to raise additional funds would be more likely to default than its observable
characteristics would otherwise suggest. Railroad management may well have been able to convey
that information to the RFC board in order to increase the likelihood that an RFC loan would be
granted. In that situation, the error from the regression of bond defaults on bailouts is likely to be
14
This finding is robust to changes in the regression specification, such as OLS fixed-effects models.

15
correlated with the independent variable Approval. Therefore, the coefficient estimates on Approval,
which measure the effectiveness of government aid, will be biased.
We would like to use an instrumental variable that is correlated with a railroad receiving a government
bailout but only affects a railroad’s financial performance via the granting of RFC loans. We take
advantage of the prior literature (see e.g., Wright (1974), Wallis (1987), and Fishback, Kantor, and
Wallis (2003)) that claims New Deal grants were influenced by politics. Fishback (2017), for
example, concludes: “Nearly every study finds that political considerations were important to the
Roosevelt administration.” There are, however, a few investigations of New Deal funding--such as
Mason (2003)--that find little political influence on the process. We use the composition of the RFC
board as our instrumental variable. Specifically, we use the number of states a railroad passed through
that were the home states of RFC directors in that particular year. We call this the number of a
railroad’s Connections to the RFC. For example, on February 5, 1932, the Chicago and Eastern
Illinois railroad applied for an RFC loan for $3.629 million. This railroad passed through Illinois,
Indiana, and Missouri. H. Paul Bestor (Missouri) and Charles G. Dawes (Illinois) sat on the board of
the RFC at the time of the application. Therefore, our instrument takes a value of two.
In our first stage regression (Table 6, columns 2 and 5) we regress Approval on a railroad’s lagged
characteristics and our instrument, Connections. We see that Approval is positively and statistically
significantly related to a railroad’s Connections, even with region and year fixed effects. The F-
statistic in the first stage regression is 75.791, which indicates that we have a strong instrument.
To have a valid instrument, we also require that the exclusion restriction is satisfied. The exclusion
restriction requires that Connections are uncorrelated with the error term, the unobservable part of a
railroad’s financial position that partly determines default behavior. There was, however, no realistic
possibility that a railroad that was doing poorly based on unobservable factors would increase its
number of Connections by altering its operations. Total track mileage in the U.S. declined from 1930
onwards, and it would be expensive and take years of construction for an existing railroad to begin
operations in the home state of an RFC director.
15
It also would invalidate our instrument if railroads that were in worse financial shape than their
observable characteristics would suggest were able to influence the president to alter the RFC board’s
composition, such that a new director was appointed from a state in which the railroad operated. RFC
directors were responsible for approving all loans that the Corporation made, railroad and non-
15
See ICC 53rd Annual Report on the Statistics of Railways in the United States (1939), Table 1-A.

16
railroad alike. Total railroad loans comprised less than 10% by value of the RFC’s disbursements,
and loans to an individual railroad were a tiny fraction of total RFC expenditure. There were only
five to seven directors at any one time, and the composition was balanced by political affiliation and
by the need to have directors come from different parts of the country. Given these constraints on the
composition of the RFC board, we believe it is extremely unlikely that certain railroads could have
increased their Connections by lobbying. Therefore, we use Connections as our instrument for RFC
bailouts.
In column 3 of Table 6, we replace Approval with the predicted level of Approval from our first-stage
regression. We see that bailouts increased, rather than decreased, the likelihood that a railroad would
default on its bond. This coefficient is statistically insignificant. We also observe that railroads that
were more profitable and had more cash were less likely to default, as expected. More bonds falling
due from 1930 to 1934 is positively associated with more defaults. Once we focus on the longer-run
impact (column 6), the estimated coefficient on government loan Approval is still positive but
statistically insignificant. Other railroad characteristics are little changed.
Overall, it is difficult to believe that receiving a government loan, all else equal, increased the
likelihood of a bond default, but we interpret our findings as a lack of evidence that RFC loans helped
railroads to avoid defaulting on their debt.
5.3 Bond ratings
Bailed-out railroads may have been viewed as “too big to fail” or perhaps investors anticipated that a
bailout indicated that the government would share the financial losses with bondholders. In this case,
the bond market may have perceived the railroad’s debt as being safer, despite our evidence in Table
6 showing that bailouts did not help a railroad to avoid default. To investigate perceptions of a
railroad’s default, we turn to an examination of bond ratings from Moody’s.
In table 7 we examine the likelihood that, conditional on receiving a government bailout (Approval)
last year and last year’s observable characteristics, if a railroad would receive a downgrade in the
current year. A railroad bond has a 12.65% lower probability of being downgraded in the current year
if the railroad was bailed out in the previous year (column 1). Bonds have a 6.52% lower probability
of receiving multiple downgrades, in year t, t+1, or t+2, if the firm was bailed out in year t-1 (column
4). Railroads with more cash and lower employment are less likely to be downgraded.
Since concerns about selection effects remain, we again run instrumental variable probit regressions.
The first stage regressions appear in columns 2 and 5, and the second stage results are in columns 3

17
and 6. The IV results confirm the probit results. Bailouts in the previous year are associated with a
decrease in the likelihood of getting one downgrade of 47.46% (column 3) or a decrease in the
likelihood of getting multiple downgrades of 87.13% in the subsequent three years (column 6).
Railroads that increased their employment or decreased their cash-to-assets ratio were more likely to
be downgraded. This result suggests that Moody’s perceived increased railroad employment during
the Great Depression was incompatible with protecting bondholders’ interests. Overall, our results
show that government bailouts did not protect railroads against default, although they did alleviate
bond ratings downgrades.
16
5.4 Operating performance, difference-in-differences
We now investigate if the RFC succeeded in their second objective, which was to “stimulate
employment.” We determine the ability of government bailout recipients to improve their economic
performance, including their employment numbers. We first conduct a difference-in-differences
approach on RFC loan recipients’ profitability, leverage, employment, and wage bill as a fraction of
total expenses. In Table 8, we use the technique of Callaway and Sant’Anna (2020) to deal with
staggered bailouts. For each bailed-out railroad’s initial government loan, we choose a matched
railroad. The matched railroad must operate in the same ICC region and have total assets that are
most similar to those of the bailed-out carrier. In addition, the matched railroad must not have received
a bailout during the period between two calendar years before the bailout and two calendar years after
the bailout. If a railroad received multiple bailouts, we also include those observations—as long as
the subsequent bailouts were at least three years after the prior bailout.
In Panel A, we present average treatment effects (ATT).
17
We find negative but statistically
insignificant effects of a bailout on railroad profits (columns 1-3). The estimated treatment effect for
leverage is a reduction of 3.3 percentage points for the bailed-out railroad (column 4). The starting
leverage for the average railroad was a little over 40 percent of total assets (see Table 1). We find
positive but economically small and statistically insignificant effects of a bailout on employment
16
In Table 6, we show that Moody’s ratings add information to understand bond default dynamics, even after controlling
for firm characteristics. However, Moody’s ability to discern between good and bad corporate bonds comes mostly from
non-investment grade bonds. Moody’s ratings add very little information for investors that is not already conveyed by
firm characteristics for investment grade bonds. The only bond rating fixed effects which are significantly different to
zero are the lowest rated C and Ca. Most government bailouts went to railroads with investment-grade bonds. Government
bailouts helped such bonds to preserve their (high) credit rating (Table 7) and since investment-grade bonds are very
unlikely to default, a bailout did not greatly change their default risk (Table 6).
17
Using a chi-squared test, we highlight that the parallel trends assumption is never violated in Table 8.

18
(columns 7-9). In contrast, we note statistically significant increases in the fraction of total expenses
going to the wage bill of around 4.5%. Our results of weak employment and generous wages align
with the findings of Ohanian (2009) and Cole and Ohanian (2004) that New Deal policies deepened
the Great Depression.
18
Railroads appear to have used some of the government funds to inflate their
wages bill.
In Panel B, we present estimates of the treatment effect by year. There are no statistically significant
impacts on profitability or employment after the treatment. In contrast, we find that leverage
decreases in years t+1 through t+4, but it is barely affected in the year of the bailout. Wages increase
in the year of the bailout and continue to increase for the following three years.
In Panel C, we run a placebo test in which we counterfactually assume that all RFC bailouts took
place in 1929, while still focusing on the “actual” bailed-out vs. non-bailed-out railroads (as in Berger
and Roman (2017)). For the placebo test, we use the years 1927 to 1928 as the “pre-RFC period,” and
the period between 1930 and 1932 as the “post-RFC period.” We apply the doubly-robust difference-
in-differences approach of Sant’Anna and Zhao (2020). We fail to find significant results for
profitability, leverage, employment, or the wage bill with our placebo.
5.5 Operating performance, instrumental variables
Although the difference-in-differences approach should give us a good idea of the impact of a railroad
bailout, there remain concerns that the comparison group of non-bailed-out railroads does not provide
an accurate counterfactual for the bailed-out carriers. In Table 9, we again make use of the board
composition of the RFC and our measure, Connections, as an instrument for railroad bailouts. In the
second stage, we regress railroad profitability, leverage, log employment, and the wage bill fraction
on the fitted level of bailouts after conditioning on railroad characteristics.
We find no statistically significant impact of railroad bailouts on profitability (column 2),
employment (column 4), or the wage bill fraction (column 5). We do find that a bailout causes an 8.8
percentage point decrease in leverage (column 3). Therefore, we conclude that the RFC failed in their
second objective, which was to promote railroad employment via their loan program. All regressions
use firm, year, and railroad region fixed effects and condition on lagged characteristics, including the
railroad’s freight composition. Our results are in line with those of Granja, Makridis, Yannelis, and
Zwick (2020) who find that small business support payments during the coronavirus pandemic were
18
Schiffman (2003) finds that railroads that defaulted increased their employment following the default.

19
often used to make non-payroll payments and to build up savings buffers. Railroads appear to have
used bailouts to reduce their leverage with no beneficial impact on employment.
5.6 Economic spillovers
Bailouts do not seem to have provided any direct benefits for the recipient carriers—save a jump in
the value of their debt. They may, however, have provided spillover benefits for the regions in which
they operated. For example, railroads may have been able to keep operating routes that would
otherwise have been closed, or they may have conducted a more frequent schedule that permitted
local businesses to operate more smoothly than if government support had not been made available.
We examine if there were positive economic spillovers that flowed from the bailouts of railways that
passed through a city. We create an explanatory variable, City RFC Approvals, which equals the
fraction of all railroads that operate in a city that received an RFC or PWA loan in the previous year.
We again use our instrumental variable, Connections, which is measured at the state level, as an
instrument. We regress the natural logarithm of city building approvals per capita in a year on fitted
City RFC Approvals.
19
In Table 10, we see that RFC board connections are very strong instruments
for city-level loan approvals. We find a negative relationship between railroad city-level loan
approvals and new building approvals (columns 2 and 4). Once we add both year and city fixed
effects, however, the estimated impact of City RFC Approvals on building approvals becomes
statistically insignificant and close to zero in magnitude (column 6).
In Table 11, we examine if news of a railroad’s bailout affected other railroads and manufacturing
firms listed on the NYSE.
20
We calculate the abnormal returns of other firms on the day of the
railroad’s approval announcement and the cumulative abnormal return from the day before to the day
after the approval for the other railroads and manufacturing firms. The NYSE hosted three main types
of firms: railroads, manufacturing firms, and utilities. Therefore, the abnormal returns essentially
measure the extent to which the railroad sector outperformed manufacturing and utilities on the day
of a railroad bailout (when we calculate the abnormal returns for non-bailed-out railroads), and the
extent to which the manufacturing sector outperformed railroads and utilities (when we calculate the
abnormal return for manufacturing firms). The more interesting evidence looks at the cross-sectional
19
Our thanks to Gustavo S. Cortes for sharing his data on building approvals.
20
We exclude all stocks that have a zero return on all days of the event study.

20
impact of bailouts: which manufacturing firms and which railroads benefited most from news of one
railroad’s bailout?
We cross-sectionally split firms on two dimensions. First, did the other railroad overlap at all with
the bailed-out railroad, meaning did both railroads service at least one common city (Yes) or not (No)?
Second, was the level of overlap (the fraction of the bailed-out railroad’s cities also serviced by the
other railroad) above the sample mean (High) or was the overlap positive but below the sample mean
(Low). We construct similar measures of overlap for the manufacturing firms, but we consider joint
presences of manufacturing establishments (as reported in Moody’s Manual of Investments –
Industrial Securities) and railroad tracks.
In Panel A, we see that the mean CAR for all types of railroads was large and positive at the time of
an application, but there was no statistical difference between railroads that overlapped with the
bailed-out railroad and non-overlapping railroads. In Panel B, we observe slightly larger, 0.5% (No
less Yes) to 0.6% (Low less High), and statistically significant differences in CARs for other railroads.
We interpret this result to mean that competing railroads (i.e., those with some overlap with the
bailout recipient) suffered from a bailout relative to railroads that had little or no overlap. As this is a
cross-sectional test, we are conditioning on any economy-wide railroad shocks such as changes in
government railroad policy, input costs, or demand changes.
Manufacturing firms experienced positive abnormal returns at the time of a railroad bailout (relative
to utilities and railroads themselves). Again, however, our interest lies in the differences between
manufacturing firms that were co-located in the same city as the bailed-out railroad (an overlap of
Yes or High) and manufacturing firms that were not located in cities through which the bailed-out
railroad ran (No or Low). We see that a railroad bailout benefited co-located manufacturing firms
relative to manufacturers that were not located near the bailout recipient’s tracks. Co-located
manufacturing firms outperformed others by 0.1% (Yes vs. No) at the time of the application and
0.2% at the time of the approval. If we compare High vs. Low manufacturing firms, we see that high-
overlap manufacturers outperformed by 0.6% at the time of application (Panel A) and 0.3% at the
time of approval (Panel B). Taken together, we feel there is some evidence that there were positive

21
economic spillovers to the real economy from railroad bailouts, even if the railroads themselves
showed little benefit.
5 Conclusion
The RFC distributed much of the U.S. government’s New Deal assistance to the economy as it
struggled with the Great Depression. Around 10% of the RFC’s loans were given to private firms in
the railroad sector. In our study, we ask if RFC assistance aided railroads in avoiding debt defaults
and maintaining employment. We find no evidence that government assistance was beneficial for the
recipient railroad and some evidence that government assistance caused harm – point estimates are
that government bailouts increased the likelihood of a debt default. A bailed-out railroad may have
felt pressure to maintain employment at higher than desired levels and/or to keep wages above market
levels as the Depression deepened.
Non-bailed-out railroads that competed with the bailout recipient seem to have suffered some harm
from the bailout, presumably because one of their competitors was supported financially. We find
some evidence that government railroad support was beneficial for manufacturing firms that were co-
located near the railroad’s tracks. Therefore, although RFC and PWA assistance proved of little
benefit to the railroad itself, there were positive economic spillovers from this New Deal program.

22
Figure 1: New railroad bond issues
Number of new bond issues by all class I railroads between 1927 and 1939
Figure 2: RFC railroad loans ($ million, including roll-overs)
Value of approved bailouts for all class I railroads between 1932 and 1939 on a quarterly basis
0
5
10
15
20
25
1927 1928 1929 1930 1931 1932 1933 1934 1935 1936 1937 1938 1939
Number of new bonds
Years
0
50
100
150
200
Q1 1932
Q2 1932
Q3 1932
Q4 1932
Q1 1933
Q2 1933
Q3 1933
Q4 1933
Q1 1934
Q2 1934
Q3 1934
Q4 1934
Q1 1935
Q2 1935
Q3 1935
Q4 1935
Q1 1936
Q2 1936
Q3 1936
Q4 1936
Q1 1937
Q2 1937
Q3 1937
Q4 1937
Q1 1938
Q2 1938
Q3 1938
Q4 1938
Q1 1939
Q2 1939
Q3 1939
Q4 1939

23
Figure 3: Number of loan approvals
The number of RFC or PWA loans to railroads that operated in each state. For railroads that operated in more than one
state, we count each state in which that railroad operated as having received a loan.
Figure 4: Kaplan-Meier failure graphs
We show the hazard rates of bond defaults in the years after a bailout. Panel A shows the hazard rates for bailed-out vs.
non-bailed-out railroads. Panel B shows the hazard rates after controlling for lagged log total assets, net income to total
assets, cash to total assets, leverage, log employment, log firm age, volatility, and the freight composition.
Panel A. Benchmark model
Panel B. Controlling for covariates

24
Table 1: RFC board composition
Panel A reports the RFC board composition between 1932 and 1939. Party refers to the respective RFC member’s political
party, where Dem. refers to Democrat and Rep. refers to Republican. State refers to the respective member’s home state
(abbreviated). Start and End refers to the member’s start and end time on the board, respectively. Comments gives insights
into their background. In Panel B, the dependent variable equals one if an RFC member from state y was appointed to the
board in year t, zero otherwise. Column 1 includes all RFC members; column 2 excludes the first RFC board; column 3
examines RFC members with previous political experience; column 4 examines RFC members from the private sector.
Bank Capital in Default is the ratio of national bank capital in default to total national bank capital in state y. Log(Size
per Railroad) is the logarithm of the total assets of railroads active in state y divided by the number of railroads active in
state y. Number of Railroads is the number of railroads active in state y. Railroad Bailout Weight is the ratio of the total
assets of bailed-out railroads in year t active in state y to the total assets of railroads active in state y. Log(Building permits
per capita) is the log of building permits divided by the state population in state y. Banks per capita is the number of
national banks in state y divided by state population. Number of RFC members is the number of existing RFC members
from state y. All regressions use year and state fixed effects. We cluster standard errors at the state level. *, **, and ***
denote significance at the 10%, 5% and 1% levels, respectively.
Name
Party
State
Start
End
Comments
H. Paul Bestor
Rep.
MO
Feb-32
Jul-32
ex officio as Farm Loan Commissioner
Eugene Meyer
Rep.
NY
Feb-32
Jul-32
ex officio as governor of Fed. Reserve
Andrew W. Mellon
Rep.
PA
Feb-32
Feb-32
ex officio as Treasury Secretary
Ogden L. Mills
Rep.
NY
Feb-32
Mar-33
ex officio as Treasury (Under) Secretary
William Woodin
Dem.
NY
Mar-33
Dec-33
ex officio as Treasury Secretary
Arthur H. Ballantine
Rep.
-
Feb-32
May-33
ex officio as Treasury Under Secretary
Dean H. Acheson
Dem.
-
May-33
Nov-33
ex officio as Treasury Under Secretary
Henry Morgenthau (Jr.)
Dem.
NY
Nov-33
Feb-38
ex officio as Treasury (Under) Secretary
Thomas J. Coolidge (III)
Dem.
MA
May-34
Feb-36
ex officio as Treasury (Under) Secretary
Roswell Maginn
Dem.
IL
Jan-37
Feb-38
ex officio as Treasury (Under) Secretary
Harvey C. Couch
Dem.
AR
Feb-32
Aug-34
Arkansas businessman (electricity, railways)
Charles G. Dawes
Rep.
IL
Feb-32
Jun-32
Former Vice President and Chicago banker
Jesse H. Jones
Dem.
TX
Feb-32
Jul-39
Texas businessman (lumber, real estate, banking)
Wilson McCarthy
Dem.
UT
Feb-32
Sep-33
Utah state senator and district attorney
Gardner Cowles (Sen.)
Rep.
IA
Jul-32
Apr-33
Des Moines newspaper proprietor
Charles A. Miller
Rep.
NY
Aug-32
Mar-33
Utica banker
Atlee Pomerene
Dem.
OH
Aug-32
Mar-33
Ohio lawyer and former U.S. senator
Carroll B. Merriam
Rep.
KS
Jun-33
Dec-41
Topeka finance industry
John J. Blaine
Rep.
WI
Jun-33
Apr-34
Lawyer and businessman, former U.S. senator
Frederic H. Taber
Rep.
MA
Jun-33
Jan-38
New Bedford lawyer
Charles B. Henderson
Dem.
NV
Feb-34
Jul-47
Former U.S. senator and lawyer
Hubert T. Stephens
Dem.
MS
Mar-35
Feb-36
Former U.S. senator and lawyer
Charles T. Fisher (Jr.)
Rep.
MI
Mar-35
Dec-36
Detroit banker
Emil Schram
Dem.
IN
Jun-36
Jul-41
Indiana farmer and irrigator
Howard J. Klossner
Rep.
MN
Apr-37
Jul-45
Minnesota banker
Sam H. Husbands
Dem.
SC
Aug-39
Jan-46
South Carolina banker

25
Panel B – Determinants of RFC Board composition
RFC Board
RFC excl. First
Political background
Business background
(1)
(2)
(3)
(4)
Bank Capital in Default
0.078
(0.530)
-0.037
(0.690)
-0.069
(0.410)
0.032
(0.326)
Banks per capita
2.577
(0.577)
1.515
(0.495)
-0.659
(0.606)
2.174*
(0.096)
Log(Building permits per capita)
-0.019
(0.682)
-0.028
(0.500)
-0.003
(0.885)
-0.025
(0.475)
Log(State Population)
0.716
(0.418)
0.879
(0.141)
0.901*
(0.061)
-0.021
(0.928)
Log(Size per Railroad)
0.773
(0.428)
-0.452
(0.310)
-0.465
(0.231)
0.013
(0.943)
Log(Number of Railroads)
0.108
(0.185)
-0.014
(0.536)
-0.024*
(0.090)
0.011
(0.507)
Railroad Bailout Weight
0.102
(0.263)
0.142
(0.122)
0,985
(0.240)
0.047
(0.207)
Number of RFC members
-0.159**
(0.013)
-0.008
(0.947)
-0.113*
(0.053)
0.105
(0.147)
R2
0.179
0.058
0.084
0.094
Observations
344
344
344
344

26
Table 2: Summary statistics
The sample consists of 1,928 annual observations for 183 railroads from 1927 to 1939. A bailout is defined as any loan
from the Reconstruction Finance Corporation or Public Works Administration from 1932 to 1939. Connections is the
number of states the railroad operated in that were homes to RFC directors in that year. Leverage is the ratio of long-term
debt to total assets. Bonds Due / T.A. is the value of all bonds due between 1930 and 1934 to total assets in 1929. Passenger
/ Total Revenue equals passenger revenue divided by total revenue. Volatility is the standard deviation of the previous 12
month’s earnings (if earnings was missing, the 12-month standard deviation of stock returns). Wage Bill is the
compensation for employees divided by total operating expenses. We report tests of differences in means (t-test) and
medians (Wilcoxon) between the groups. *, **, and *** denote significance at the 10%, 5% and 1% levels, respectively.
Panel A: Full sample
Bailed-out Railroads
(N = 585 Railroad-Years)
Non-Bailed-Out Railroads
(N = 1,343 Railroad-Years)
Difference
Mean Median
Mean
Median
Mean
Median
T-test
Wilcoxon
Log (Age, years)
3.533
3.638
3.499
3.611
0.033
0.027*
Cash / T.A.
0.015
0.012
0.024
0.014
-0.009***
-0.002**
Connections
1.481
1.000
0.745
1.000
0.735***
0.000
Log (Employment)
9.185
9.454
7.126
7.012
2.059***
2.442***
Leverage
0.416
0.412
0.439
0.406
-0.023*
0.006
Net income / T.A.
0.009
0.012
0.016
0.010
-0.006**
0.002
Passenger / Total Revenue
0.110
0.959
0.099
0.059
0.011*
0.041***
Log (Total Assets)
19.300
19.379
17.116
16.938
2.189***
2.441***
Volatility
0.141
0.091
0.214
0.131
-0.074***
-0.040***
Wage Bill
0.711
0.636
0.626
0.616
0.086*
0.020***
Panel B: 1929 railroad characteristics
Bailed-out Railroads
(N = 46 Railroads)
Non-Bailed-Out Railroads
(N = 118 Railroads)
Difference
Mean Median
Mean
Median
Mean
Median
T-test
Wilcoxon
Log (Age, years)
3.364
3.526
3.347
3.481
0.017
0.045
Cash / T.A.
0.018
0.014
0.028
0.017
-0.009*
-0.003
Bonds Due1930-1934 / T.A.
0.041
0.011
0.022
0.000
0.019
0.011***
Connections1932
1.087
1.000
0.496
0.000
0.591***
1.000****
Log (Employment)
9.491
9.762
7.430
7.372
2.061***
2.390***
Leverage
0.418
0.402
0.418
0.389
0.000
0.013
Net income / T.A.
0.027
0.026
0.030
0.027
-0.004
-0.001
Passenger / Total Revenue
0.096
0.079
0.075
0.039
0.021
0.040***
Log (Total Assets)
19.297
19.431
17.174
16.972
2.122***
2.479***
Volatility
0.077
0.063
0.119
0.084
-0.043***
-0.021***
Wage Bill
0.643
0.644
0.716
0.623
-0.074
0.021
Panel C: Full sample
Multiple Bailouts
(N = 485 Railroad-Years)
One Bailout
(N = 100 Railroad-Years)
Difference
Mean Median
Mean
Median
Mean
Median
T-test
Wilcoxon
Log (Age, years)
3.467
3.611
3.880
4.060
-0.413***
-0.449***
Cash / T.A.
0.014
0.012
0.018
0.013
-0.004***
-0.001
Connections
1.662
1.000
0.736
1.000
0.926***
0.000
Log (Employment)
9.306
9.504
8.533
8.342
0.772***
1.162***
Leverage
0.418
0.419
0.403
0.389
0.015
0.030***
Net income / T.A.
0.008
0.010
0.016
0.019
-0.007**
-0.009***
Passenger / Total Revenue
0.115
0.099
0.085
0.081
0.030***
0.018***
Log (Total Assets)
19.428
19.484
18.609
18.654
0.819***
0.830***
Volatility
0.138
0.090
0.157
0.095
-0.019
-0.005
Wage Bill
0.726
0.638
0.069
0.641
0.633
-0.003

27
Table 3: Determinants of bailout
We regress bailouts on railroad characteristics. Column 1 presents the first-stage probit coefficients. The dependent
variable, Application, equals one if the railroad applied for at least one loan in that year, and zero otherwise. Columns 2
to 4 present the second-stage regressions where the dependent variable¸ Approval, equals one if the railroad got at least
one application approved, and zero otherwise. We use a probit (column 2), OLS (column 3), and the Heckman Selection
Model (column 4) to calculate the second-stage regression coefficients. Approval (in last 3 Years) is a dummy variable
that equals one if the railroad had an RFC or PWA loan approved in the last three years, and zero otherwise. Cum. Loan
Size / Total Assets equals the cumulative bailout loan amount a railroad has received since 1932 divided by its total assets.
Other variables are as defined in Table 2. *, **, and *** denote significance at the 10%, 5% and 1% levels, respectively.
First Stage
Second Stage
(Application)
(Approval)
(1)
(2)
(3)
(4)
Connections
0.100
(0.148)
0.975***
(0.000)
1.114***
(0.000)
1.013***
(0.000)
Log (Total Assets)
0.358**
(0.017)
0.012
(0.220)
0.013
(0.204)
0.012
(0.234)
Net income / T.A.
-0.965
(0.763)
0.099
(0.315)
0.156
(0.202)
0.121
(0.473)
Leverage
0.771
(0.106)
-0.003
(0.870)
0.000
(0.987)
0.001
(0.967)
Cash / T. A.
-15.964**
(0.038)
0.199
(0.118)
0.308*
(0.087)
0.376
(0.229)
Log (Age, years)
-0.231**
(0.029)
-0.004
(0.675)
-0.007
(0.471)
-0.006
(0.510)
Volatility
0.159
(0.500)
-0.005
(0.875)
-0.006
(0.692)
-0.005
(0.842)
Log (Employment)
0.165
(0.284)
-0.004
(0.679)
-0.003
(0.809)
-0.004
(0.728)
Passenger / Total Revenue
-2.404*
(0.062)
0.033
(0.295)
0.045
(0.202)
0.043
(0.374)
Bonds Due1930-1934 / T.A.
-0.798
(0.661)
0.059
(0.563)
0.047
(0.677)
0.054
(0.626)
Approval (in Last 3 Years)
0.159
(0.328)
-0.389*
(0.078)
-0.065
(0.232)
Cum. Loan size / Total Assets
0.778
(0.625)
2.706
(0.203)
0.0626
(0.177)
Pseudo R2 / R2
0.417
0.345
0.214
0.458
Year FE
Yes
Yes
Yes
Yes
Region FE
Yes
Yes
Yes
Yes
Freight Composition
Yes
Yes
Yes
Yes
Specification
Probit
Probit
OLS
Heckman
Observations
1,554
1,548
1,548
1,548

28
Table 4: Announcement effects
We calculate the abnormal returns (AR) and cumulative abnormal returns (CAR) of a security from four days before to
four days after the announcement of an application, approval, or refusal. We measure abnormal returns as the security’s
returns less the Moody’s bond index / CRSP market index on the same day. We average the AR (CAR) across securities.
Standard errors are clustered by railroads. p-values appear in parentheses. Panel A presents the results for all bailouts.
Panel B presents the results for the initial and subsequent bailouts. *, **, and *** denote significance at the 10%, 5% and
1% levels, respectively.
Panel A: All bailouts
Applications
Approvals
Refusals
Bonds
Equity
Bonds
Equity
Bonds
Equity
Day -4
-0.002
(0.275)
-0.007
(0.237)
0.001
(0.384)
0.005
(0.289)
-0.004
(0.526)
0.048
(0.320)
Day -3
0.007
(0.414)
0.009
(0.181)
0.002
(0.249)
-0.003
(0.548)
0.005
(0.757)
0.002
(0.916)
Day -2
0.000
(0.993)
0.002
(0.681)
-0.000
(0.892)
0.005
(0.397)
-0.005
(0.657)
-0.032
(0.174)
Day -1
0.006
(0.296)
-0.003
(0.969)
-0.002
(0.529)
0.006
(0.419)
0.000
(0.994)
-0.079*
(0.092)
Day 0
0.009**
(0.049)
0.004
(0.550)
0.003*
(0.092)
-0.025
(0.516)
0.016
(0.646)
0.044
(0.443)
Day +1
0.003*
(0.097)
0.006
(0.302)
0.004**
(0.019)
-0.008
(0.161)
-0.059**
(0.047)
-0.061*
(0.100)
Day +2
0.015***
(0.008)
-0.001
(0.922)
0.002
(0.188)
0.002
(0.698)
-0.013
(0.548)
0.027
(0.121)
Day +3
0.003
(0.146)
0.003
(0.646)
0.005**
(0.023)
-0.009
(0.175)
0.009
(0.351)
-0.029
(0.337)
Day +4
0.001
(0.615)
-0.005
(0.263)
0.000
(0.830)
0.007
(0.215)
0.019*
(0.100)
0.005
(0.749)
CAR (t-4, t+4)
0.042***
(0.001)
0.012
(0.281)
0.016***
(0.009)
-0.019
(0.618)
-0.025
(0.971)
-0.075
(0.247)
Observations
344
134
492
190
19
9
Panel B: First vs. subsequent bailouts
First bailout
Subsequent bailouts
Applications
Approval
Applications
Approval
Bonds
Equity
Bonds
Equity
Bonds
Equity
Bonds
Equity
Day -4
-0.002
(0.577)
0.004
(0.731)
-0.007*
(0.054)
0.001
(0.942)
-0.002
(0.348)
-0.010
(0.159)
0.003*
(0.079)
0.006
(0.248)
Day -3
0.020
(0.191)
0.029**
(0.250)
-0.001
(0.818)
0.013
(0.537)
-0.001
(0.793)
0.004
(0.631)
0.002
(0.209)
-0.007
(0.221)
Day -2
0.001
(0.946)
-0.033***
(0.001)
0.004
(0.643)
0.003
(0.862)
-0.000
(0.900)
0.013*
(0.069)
-0.001
(0.558)
0.006
(0.394)
Day -1
0.008
(0.424)
0.004
(0.785)
-0.001
(0.864)
-0.012
(0.575)
0.016
(0.304)
-0.001
(0.860)
-0.002
(0.545)
0.009
(0.248)
Day 0
0.026
(0.171)
0.001
(0.931)
0.009**
(0.043)
-0.229
(0.317)
0.004**
(0.023)
0.005
(0.505)
0.002
(0.351)
0.015*
(0.089)
Day +1
0.004
(0.372)
-0.007
(0.591)
0.005
(0.206)
0.001
(0.974)
0.003
(0.163)
0.011
(0.125)
0.004**
(0.042)
-0.010*
(0.084)
Day +2
0.035*
(0.074)
0.029***
(0.008)
0.004
(0.259)
0.011
(0.461)
0.008**
(0.029)
-0.009
(0.211)
0.002
(0.319)
0.003
(0.995)
Day +3
0.000
(0.974)
-0.009
(0.559)
0.021**
(0.040)
-0.037*
(0.093)
0.004
(0.042)
0.007
(0.307)
0.002
(0.291)
-0.003
(0.618)
Day +4
0.005
(0.466)
-0.001
(0.861)
0.007**
(0.027)
0.012
(0.472)
-0.000
(0.962)
-0.006
(0.302)
-0.001
(0.578)
0.006
(0.309)
CAR (t-4, t+4)
0.098**
(0.042)
0.017
(0.563)
0.040**
(0.015)
-0.239
(0.302)
0.024***
(0.008)
0.013
(0.362)
0.011*
(0.092)
0.023*
(0.081)
Observations
85
32
81
32
259
103
441
159

29
Table 5: Determinants of announcement CARs
We regress a railroad’s cumulative abnormal bond/equity return (CAR) from four days before to four days after an
application/approval of an RFC or PWA loan. Variables are as defined in Table 2. Close to default equals 1 if the railroad’s
bond price is below 50 and zero otherwise. We add firm and region fixed effects and cluster standard errors at the railroad
level. In Panel A, we focus on all loan approvals and applications. In Panel B, we focus on all first and subsequent loan
approvals and applications. *, **, and *** denote significance at the 10%, 5% and 1% levels, respectively.
Panel A: All observations
Application
Approval
Security
Bond
Equity
Bond
Equity
(1)
(2)
(3)
(4)
Connections
-0.017
(0.219)
0.005
(0.725)
0.009
(0.268)
0.064
(0.304)
Log (Total Assets)
-0.023
(0.717)
2.133
(0.161)
-0.009
(0.267)
2.755
(0.586)
Net income / T.A.
-0.321
(0.746)
1.884
(0.176)
-0.420
(0.485)
-7.550
(0.326)
Leverage
0.052
(0.790)
-2.164***
(0.003)
0.381
(0.289)
8.369
(0.229)
Cash / T.A.
-0.659
(0.801)
0.638
(0.877)
-3.622
(0.251)
6.077
(0.735)
Log (Age, years)
0.002***
(0.008)
0.006
(0.717)
0.092
(0.737)
-1.285
(0.570)
Volatility
-0.029
(0.556)
0.042
(0.465)
-0.239
(0.260)
-1.894
(0.105)
Log (Employment)
0.103
(0.164)
-1.922***
(0.005)
-0.026**
(0.015)
-1.339*
(0.068)
Passenger / Total Revenue
-0.113
(0.265)
0.085
(0.423)
-0.097
(0.151)
-0.094
(0.781)
Close to default
-0.030
(0.581)
-0.035
(0.428)
0.011
(0.421)
0.049
(0.663)
Loan size / Total Assets
0.105
(0.514)
-0.036
(0.537)
Region FE
Yes
Yes
Yes
Yes
Freight composition
Yes
Yes
Yes
Yes
Observations
313
141
411
169
R2
0.009
0.062
0.057
0.053

30
Panel B: First vs. subsequent bailouts
Application
Approval
First Bailout
Subsequent Bailouts
First Bailout
Subsequent Bailouts
Security
Bond
Equity
Bond
Equity
Bond
Equity
Bond
Equity
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Connections
-0.194
(0.162)
0.031
(0.456)
-0.000
(0.975)
0.018
(0.295)
0.094*
(0.100)
1.166*
(0.054)
0.012
(0.153)
0.006
(0.634)
Log (Total Assets)
-0.829
(0.217)
0.223
(0.816)
-0.133**
(0.034)
0.238
(0.380)
-0.867**
(0.002)
-6.513
(0.171)
-0.054
(0.650)
0.416***
(0.001)
Net income / T.A.
-1.484
(0.788)
6.533
(0.453)
-0.439
(0.615)
1.116*
(0.090)
-5.976*
(0.058)
-106.682**
(0.042)
0.249
(0.673)
0.738
(0.579)
Leverage
-1.017*
(0.056)
-0.479
(0.592)
-0.139
(0.352)
-0.361
(0.215)
-0.003
(0.993)
-1.294
(0.793)
0.277
(0.175)
-0.009
(0.979)
Cash / T.A.
-13.719
(0.503)
-22.862
(0.246)
-1.999
(0.426)
7.736
(0.173)
-14.464
(0.109)
-107.918
(0.420)
-3.821
(0.126)
5.933
(0.125)
Log (Age, years)
0.000
(0.895)
0.002
(0.635)
0.001***
(0.009)
-0.001**
(0.043)
0.030
(0.656)
-0.848
(0.411)
0.103
(0.240)
0.145**
(0.012)
Volatility
0.069
(0.310)
-0.029
(0.847)
0.076*
(0.078)
0.074
(0.668)
0.652**
(0.016)
5.702
(0.148)
-0.014
(0.897)
-0.343***
(0.004)
Log (Employment)
1.549*
(0.092)
-0.298
(0.731)
0.141**
(0.043)
-0.316
(0.175)
1.711
(0.130)
22.009**
(0.015)
-0.036***
(0.000)
0.161
(0.714)
Passenger / Total Revenue
-1.798
(0.182)
-0.470
(0.349)
0.022
(0.748)
-0.070
(0.493)
-0.756
(0.152)
-3.290
(0.645)
-0.133***
(0.007)
0.062
(0.463)
Close to default
-0.142
(0.329)
0.118
(0.528)
0.007
(0.747)
0.008
(0.848)
-0.057
(0.245)
-1.981*
(0.073)
0.010
(0.545)
0.007
(0.849)
Loan size / Total Assets
-0.071
(0.166)
-0.298
(0.388)
-0.014*
(0.089)
-0.009
(0.490)
Firm FE
No
No
Yes
Yes
No
No
Yes
Yes
Region FE
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Freight Composition
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Observations
72
31
241
102
67
28
337
141
R2
-0.012
0.154
0.059
0.058
0.333
0.151
0.002
0.081

31
Table 6: Determinants of bond defaults
We regress bond defaults on lagged railroad characteristics. Default equals one if the railroad failed to meet a coupon or
principal repayment, or in any way changed the terms of the issue in the current year. We drop all observations of the
respective railroads the year after Default equals one. Approval equals one if the railroad obtained an RFC or PWA loan
in the previous year. In columns 1 and 4, we run a probit regression model. In column 2 and 5, we present our first-stage
regression for the instrumental-variable (IV) approach. We regress an indicator variable equal to one in the year the
railroad received an Approval, and zero otherwise, on railroad controls. In columns 3 and 6, we present the second-stage
instrumental variable (IV) regression. p-values, in parentheses, are adjusted for heteroskedasticity and clustered at the
railroad-level. We include region, year and (in columns 4-6) bond rating fixed effects. For a railroad with multiple
outstanding bonds, we use the rating of the bond closest to maturity. *, **, and *** denote significance at the 10%, 5%
and 1% levels, respectively.
Bond default
(this year)
Bond default
(this year or next two years)
First Stage
Second Stage
First Stage
Second Stage
(1)
(2)
(3)
(4)
(5)
(6)
Approval
0.247
(0.531)
0.893
(0.697)
0.694**
(0.029)
1.904
(0.291)
Log (Total Assets)
0.288
(0.370)
0.028*
(0.068)
0.274
(0.412)
0.343
(0.185)
0.016
(0.181)
0.316
(0.258)
Net income / T.A.
-9.198**
(0.034)
-0.258
(0.108)
-8.857*
(0.053)
-6.259
(0.133)
-0.177
(0.202)
-5.764
(0.197)
Leverage
0.399
(0.426)
0.050
(0.161)
0.364
(0.471)
0.589
(0.251)
0.053
(0.129)
0.513
(0.271)
Cash / T.A.
-88.092***
(0.001)
0.351
(0.216)
-87.829***
(0.001)
-45.114***
(0.005)
0.183
(0.354)
-44.265***
(0.007)
Log (Age, years)
-0.480**
(0.015)
-0.051***
(0.002)
-0.447**
(0.041)
-0.397**
(0.018)
-0.024**
(0.032)
-0.362**
(0.041
Volatility
0.191
(0.525)
-0.055**
(0.015)
0.206
(0.511)
0.214
(0.281)
-0.003
(0.943)
0.205
(0.328)
Log (Employment)
-0.319
(0.314)
0.027*
(0.072)
-0.351
(0.254)
-0.407
(0.108)
0.021*
(0.088)
-0.437*
(0.064)
Bonds Due1930-1934 / T.A.
0.415**
(0.021)
-0.003
(0.711)
0.456**
(0.018)
0.323**
(0.039)
-0.000
(0.984)
0.317**
(0.027)
Passenger / Total Revenue
-3.143
(0.251)
-0.044
(0.541)
-2.935
(0.328)
-1.554
(0.439)
-0.088
(0.193)
-1.283
(0.508)
Cum. Loan Size / Total Assets
-0.968
(0.827)
8.462***
(0.000)
-6.831
(0.762)
-1.261
(0.702)
8.611***
(0.000)
-12.401
(0.440)
Connections
0.139***
(0.000)
0.147***
(0.000)
Region FE
Yes
Yes
Yes
Yes
Yes
Yes
Year FE
Yes
Yes
Yes
Yes
Yes
Yes
Rating FE
Yes
Yes
Yes
Yes
Yes
Yes
Freight Composition
Yes
Yes
Yes
Yes
Yes
Yes
Log Likelihood
-64.834
-91.317
-91.317
-185.491
-137.569
-137.569
Observations
769
769
769
1,114
1,114
1,114
F-Statistic
n.a.
69.758
n.a.
n.a.
69.758
n.a
Specification
Probit
IV-Probit
IV-Probit
Probit
IV-Probit
IV-Probit

32
Table 7: Rating changes
We regress bond rating changes, for the nearest-to-maturity bond, on lagged railroad and bond characteristics. Columns
1 and 4 contain probit regressions. The dependent variable equals one if there was a Moody’s rating downgrade in the
current year, and zero otherwise (Column 1) or if there was more than one rating downgrade from the current year to year
t+2 (Column 4). Variables are as defined in Tables 2 and 6. We include Time to maturity, the log of the number of years
to maturity for the respective bond, and the Nominal outstanding amount of the bond to total long-term debt. In columns
2 and 5, we report first-stage regressions. We regress an indicator variable equal to one the year the railroad received an
Approval, and zero otherwise, on railroad and bond controls. We present the second-stage instrumental variable (IV)
probit regressions for one downgrade (column 3) and multiple downgrades (column 6). p-values, in parentheses, are
adjusted for heteroskedasticity and clustered at the railroad-level. We include region, firm, and rating fixed effects. *, **,
and *** denote significance at the 10%, 5% and 1% levels, respectively.
Single Rating Downgrade
(this year)
Multiple Rating Downgrades
(this year or next two years)
First Stage
Second Stage
First Stage
Second Stage
(1)
(2)
(3)
(4)
(5)
(6)
Approval
-0.416***
(0.000)
-1.434***
(0.000)
-0.305***
(0.004)
-2.633***
(0.000)
Connections
0.096***
(0.000)
0.093***
(0.000)
Firm control variables
Log (Total Assets)
-0.139
(0.936)
0.262
(0.102)
0.268
(0.806)
-5.536
(0.119)
0.209
(0.130)
-1.386
(0.420)
Net income / T.A.
6.355**
(0.012)
-2.139***
(0.001)
3.110
(0.177)
1.349
(0.695)
0.034
(0.875)
-0.067
(0.952)
Leverage
0.288
(0.789)
0.769***
(0.000)
1.103
(0.269)
1.556
(0.169)
0.657***
(0.000)
2.435***
(0.000)
Cash / T.A.
-26.299***
(0.006)
2.899**
(0.022)
-20.698**
(0.017)
-22.809**
(0.045)
-0.275
(0.751)
-10.675*
(0.062)
Log (Age, years)
-1.190***
(0.006)
0.253***
(0.010)
-0.628
(0.181)
-5.495***
(0.000)
0.194**
(0.037)
-1.704***
(0.001)
Volatility
-0.256*
(0.066)
0.073***
(0.000)
-0.163
(0.213)
-0.519**
(0.018)
0.061***
(0.001)
-0.147
(0.325)
Log (Employment)
2.122***
(0.000)
-0.253**
(0.014)
1.595***
(0.003)
3.309***
(0.000)
-0.231**
(0.016)
0.831*
(0.052)
Passenger / Total Revenue
6.327**
(0.046)
-17.797
(0.273)
2.928
(0.179)
14.387***
(0.000)
-1.901
(0.207)
-0.225
(0.954)
Cum. Loan Size / T.A.
-0.567
(0.562)
1.259***
(0.003)
1.074
(0.306)
-7.288***
(0.000)
1.185***
(0.004)
0.284
(0.848)
Bond control variables
Time to Maturity
-0.068
(0.104)
0.013*
(0.085)
-0.054
(0.167)
-0.057
(0.348)
0.011
(0.119)
-0.009
(0.781)
Nominal Outstanding / L.T.D.
0.127**
(0.023)
-0.011
(0.204)
0.103**
(0.038)
0.128***
(0.000)
0.013**
(0.002)
0.086***
(0.000)
Region FE
Yes
Yes
Yes
Yes
Yes
Yes
Firm FE
Yes
Yes
Yes
Yes
Yes
Yes
Rating FE
Yes
Yes
Yes
Yes
Yes
Yes
Freight Composition
Yes
Yes
Yes
Yes
Yes
Yes
Log likelihood
-950.939
-1,546.654
-1,546.654
-757,378
-1,356.539
-1,356.539
Observations
1,779
1,779
1,779
1,992
1,992
1,992
F-Statistic
n.a.
118.461
n.a.
n.a.
118.461
n.a.
Specification
Probit
IV-Probit
IV-Probit
Probit
IV-Probit
IV-Probit

33
Table 8: Difference-in-difference: Profitability, leverage, employment, and wage bill
We present difference-in-difference regressions for profitability, leverage, and employment with variation in treatment timing and multiple time periods following Callaway
and Sant’Anna (2020). In Panel A, ATT is defined as the average treatment effect for the treated subpopulation. Panel B presents results for an event study analysis. Panel C
presents results if we assume that bailed-out railroads (counterfactually) received an RFC or PWA loan in 1929, and the post-bailout-period was 1929-32. Columns 1, 4, 7, and
10 include all bailed-out railroads. Columns 2, 5, 8, and 11 include only those railroads that received a single bailout, whereas columns 3, 6, 9, and 12 include only those
railroads that received multiple bailouts. The p-values in parentheses use doubly robust standard errors, following Sant’Anna and Zhao (2020). All regressions use year and
region fixed effects. *, **, and *** denote significance at the 10%, 5% and 1% levels, respectively.
Panel A: Average treatment
Profitability
Leverage
Employment
Wage Bill
All Bailouts
1 Bailout
Multiple
All Bailouts
1 Bailout
Multiple
All Bailouts
1 Bailout
Multiple
All Bailouts
1 Bailout
Multiple
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
(9)
(10)
(11)
(12)
ATT
-0.002
(0.640)
-0.006
(0.241)
-0.002
(0.723)
-0.033**
(0.032)
-0.050
(0.227)
-0.032*
(0.053)
0.019
(0.597)
0.102
(0.614)
0.010
(0.764)
0.045*
(0.100)
0.011
(0.712)
0.049*
(0.098)
Obs.
1,838
1,346
1,747
1,838
1,346
1,747
1,838
1,346
1,747
1,921
1,429
1,830
Panel B: Event study
Year - 4
-0.002
(0.479)
-0.000
(0.923)
-0.003
(0.481)
0.007
(0.295)
0.012
(0.556)
0.006
(0.399)
0.073
(0.193)
-0.014
(0.483)
0.089
(0.176)
0.060
(0.771)
-0.101
(0.353)
0.089
(0.712)
Year - 3
0.000
(0.923)
-0.006*
(0.100)
0.001
(0.666)
0.006
(0.367)
0.008
(0.518)
0.006
(0.427)
-0.004
(0.703)
-0.023
(0.476)
-0.001
(0.949)
-0.185
(0.395)
-0.015
(0.405)
-0.221
(0.388)
Year -2
-0.001
(0.766)
-0.003
(0.447)
-0.000
(0.902)
0.004
(0.564)
-0.019
(0.209)
0.008
(0.261)
0.001
(0.959)
0.007
(0.765)
-0.000
(0.981)
0.123
(0.325)
-0.002
(0.813)
0.146
(0.323)
Year -1
-0.001
(0.494)
0.010*
(0.098)
-0.004
(0.175)
-0.021**
(0.017)
-0.008
(0.235)
-0.024**
(0.022)
-0.003
(0.783)
-0.023
(0.450)
0.001
(0.968)
-0.022
(0.276)
-0.005
(0.436)
-0.025
(0.291)
Year 0
-0.004
(0.156)
-0.006
(0.194)
-0.004
(0.247)
0.000
(0.987)
0.006
(0.781)
-0.001
(0.859)
0.012
(0.386)
0.006
(0.789)
0.013
(0.406)
0.036*
(0.085)
0.018
(0.152)
0.039
(0.110)
Year +1
0.010
(0.748
-0.010
(0.357)
0.003
(0.531)
-0.021*
(0.057)
-0.003
(0.755)
-0.0237*
(0.046)
0.031
(0.134)
0.012
(0.853)
0.034
(0.122)
0.044**
(0.048)
0.034
(0.146)
0.046*
(0.070)
Year +2
-0.005
(0.260)
-0.010
(0.321)
-0.004
(0.378)
-0.029*
(0.086)
-0.045
(0.464)
-0.027
(0.110)
0.042
(0.202)
0.177
(0.258)
0.027
(0.363)
0.049*
(0.061)
0.042
(0.143)
0.051*
(0.083)
Year +3
-0.015
(0.379)
-0.001
(0.912)
-0.016
(0.381)
-0.050**
(0.014)
-0.084*
(0.075)
-0.046**
(0.027)
0.014
(0.728)
0.106
(0.568)
0.003
(0.936)
0.051*
(0.058)
0.031
(0.255)
0.053*
(0.073)
Year +4
0.001
(0.848)
-0.001
(0.950)
0.001
(0.840)
-0.037*
(0.081)
-0.094
(0.132)
-0.032
(0.138)
0.022
(0.632)
0.075
(0.837)
0.017
(0.0650)
0.043
(0.136)
-0.016
(0.808)
0.048
(0.115)
Obs.
1,928
1,346
1,747
1,838
1,346
1,747
1,930
1,346
1,747
1,921
1,429
1,830
Panel C: Placebo
ATT
-0.008
(0.194)
0.006
(0.671)
0.027
(0.625)
-0.001
(0.827)
Obs.
913
913
913
913

34
Table 9: Instrumental variable regressions: Profitability, leverage, and employment
In the first stage, we regress Approval on Connections and lagged railroad characteristics. In the second stage, we regress
contemporaneous railroad profitability, leverage, employment, and the wage bill on the fitted level of lagged Approval
and lagged characteristics. Variables are as defined in Table 2 and 6. p-values are adjusted for heteroskedasticity and
clustered at the firm-level, in parentheses. *, **, and *** denote significance at the 10%, 5% and 1% levels, respectively.
First-Stage
Second-Stage
Profitability
Leverage
Employment
Wage Bill
(1)
(2)
(3)
(4)
(5)
Approval
0.008
(0.648)
-0.088*
(0.053)
-0.008
(0.940)
-0.966
(0.210)
Log (Total Assets)
0.095
(0.106)
0.007
(0.248)
0.046*
(0.052)
0.026
(0.357)
0.011
(0.864)
Cash / T. A.
0.121
(0.751)
-0.009
(0.877)
0.068
(0.593)
0.222
(0.476)
0.589
(0.520)
Log (Age, years)
-0.006
(0.937)
0.003
(0.465)
-0.007
(0.575)
0.107***
(0.005)
0.368
(0.243)
Volatility
0.017
(0.604)
-0.004*
(0.072)
-0.003
(0.521)
-0.011
(0.476)
0.062
(0.279)
Net income / T.A.
0.233
(0.154)
0.407***
(0.002)
0.006
(0.936)
-0.182
(0.357)
1.379
(0.342)
Leverage
0.007
(0.905)
0.021**
(0.045)
0.627***
(0.000)
0.089
(0.357)
-0.042
(0.734)
Log (Employment)
-0.002
(0.963)
-0.001
(0.778)
-0.008
(0.674)
0.434***
(0.000)
0.097
(0.103)
Passenger / Total Revenue
0.22
(0.961)
0.008
(0.458)
-0.016
(0.821)
-0.012
(0.887)
-4.822
(0.311)
Connections
0.066***
(0.000)
Firm FE
Yes
Yes
Yes
Yes
Yes
Year FE
Yes
Yes
Yes
Yes
Yes
Region FE
Yes
Yes
Yes
Yes
Yes
Freight Composition
Yes
Yes
Yes
Yes
Yes
R2
0.133
0.318
0.343
0.661
0.299
F statistic
76.140
n.a.
n.a.
n.a.
n.a.
Observations
1,568
1,515
1,515
1,515
1,517

35
Table 10: Railroad bailouts and building approvals
We regress the logarithm of building permits per city on City RFC Approvals (the fraction of all railroads that pass through
the city that received an RFC/PWA railroad loan approval the previous year). We condition on state-level bank
characteristics: the logarithm of bank loans per capita; the logarithm of bank deposits per capita; the logarithm of the
number of all banks; and the capital of nationally chartered banks that operated in the state that were liquidated in year t
divided by the capital of all nationally-chartered banks in that state in year t. We instrument City RFC Approvals with
Connections. p-values are adjusted for heteroskedasticity and clustered at the city-level, in parentheses. *, **, and ***
denote significance at the 10%, 5% and 1% levels, respectively.
First Stage
Second Stage
First Stage
Second Stage
First Stage
Second Stage
(1)
(2)
(3)
(4)
(5)
(6)
City RFC Approvals
-0.881***
(0.000)
-0.884***
(0.000
-0.156
(0.160)
Log (Loans per Capita)
-0.297***
(0.000)
0.188***
(0.003)
-0.314***
(0.000)
0.229***
(0.006)
-0.215
(0.225)
-0.262*
(0.057)
Log (Deposits per Capita)
-1.099***
(0.000)
0.908***
(0.000)
-2.386***
(0.000)
1.187***
(0.000)
0.557**
(0.016)
0.398*
(0.100)
Log (Number of Banks)
0.560***
(0.000)
-0.449***
(0.000)
1.501***
(0.000)
-0.748***
(0.007)
0.124
(0.524)
0.169
(0.373)
Capital of Suspended Banks
1.609**
(0.036)
1.979**
(0.018)
2.599***
(0.008)
1.851**
(0.037)
3.068***
(0.002)
1.118*
(0.057)
Connections
0.452***
(0.000)
0.412***
(0.000)
0.239***
(0.000)
Year FE
No
No
No
No
Yes
Yes
City FE
No
No
Yes
Yes
Yes
Yes
R2
0.098
0.218
0.162
0.218
0.378
0.649
Observations
2,232
2,232
2,232
2,232
2,232
2,232
F-statistic
136.0
n.a.
94.9
n.a.
39.1
n.a.

36
Table 11: Announcement effects for related firms
We calculate the abnormal return (AR) and cumulative abnormal return (CAR) for related firms’ equity after the
announcement of a railroad’s bailout application (Panel A) and approval (Panel B). We measure the AR as a firm’s equity
return less the CRSP market index. An overlap of Yes indicates the railroad/manufacturing firm operates in at least one
city with the bailed-out railroad. An overlap of No indicates the railroad/manufacturing firm does not operate in any cities
in which the bailed-out railroad operates. High indicates that the percentage overlap is above the mean level across all
firms. Low indicates that the percentage overlap is non-zero and below the mean overlap across all firms. For railroads,
the percentage overlap is defined as the number of cities that both railroads serve divided by the total number of cities of
the bailed-out railroad. For manufacturing firms, the percentage overlap is defined as the number of cities that the railroad
and manufacturing firm both operate in divided by the total number of cities the manufacturer operates in. p-values appear
in parentheses. We report the p-values of t-test differences between the groups (Yes – No or High - Low) in Diff. *, **,
and *** denote significance at the 10%, 5% and 1% levels, respectively. The returns are winsorized at the 2.5% level.
Railroads
Manufacturing
Overlap
Yes
No
High
Low
Yes
No
High
Low
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Panel A: Application
Day -1
0.003**
(0.046)
0.002**
(0.048)
0.004**
(0.043)
0.002**
(0.040)
0.003***
(0.000)
0.002***
(0.000)
0.004***
(0.001)
0.002***
(0.000)
Day 0
0.005***
(0.004)
0.003**
(0.027)
0.006**
(0.016)
0.003***
(0.009)
0.000
(0.745)
-0.004**
(0.024)
0.000
(0.781)
0.000
(0.854)
Day +1
0.006***
(0.000)
0.008***
(0.000)
0.006***
(0.009)
0.008***
(0.000)
0.003***
(0.000)
0.001***
(0.000)
0.005***
(0.000)
0.002***
(0.002)
CAR (t-1, t+1)
0.015***
(0.000)
0.013***
(0.000)
0.016***
(0.000)
0.013***
(0.000)
0.006***
(0.000)
0.005***
(0.000)
0.009***
(0.000)
0.004***
(0.000)
Diff (t-1, t+1)
0.002
[0.389]
0.003
[0.797]
0.001*
[1.769]
0.006***
[3.804]
Observations
2,208
4,861
1,442
756
12,694
68,865
4,936
7,759
Panel B: Approval
Day -1
-0.001
(0.427)
-0.000
(0.497)
-0.001
(0.949)
-0.001
(0.308)
0.003***
(0.000)
0.001***
(0.000)
0.003***
(0.000)
0.002*
(0.000)
Day 0
0.002
(0.266)
0.004***
(0.003)
0.002
(0.408)
0.003***
(0.003)
0.001***
(0.007)
0.002***
(0.000)
0.002**
(0.029)
0.001
(0.101)
Day +1
0.000
(0.756)
0.003***
(0.009)
-0.002
(0.174)
0.003***
(0.002)
0.001***
(0.005)
-0.001***
(0.008)
0.002**
(0.010)
0.007
(0.168)
CAR (t-1, t+1)
0.001
(0.545)
0.006***
(0.005)
-0.001
(0.782)
0.005***
(0.001)
0.005***
(0.000)
0.004***
(0.000)
0.007***
(0.000)
0.004***
(0.000)
Diff (t-1, t+1)
-0.005*
[-1.688]
-0.006**
[-2.053]
0.002**
[2.413]
0.003**
[2.436]
Observations
2,905
6,146
1,889
1,001
17,475
70,600
6,705
10,770

37
References
Acharya, Viral, Itamar Drechsler, and Philipp Schnabl, 2014, A Pyrrhic Victory? Bank Bailouts and Sovereign
Credit Risk, The Journal of Finance, 69(6) pp. 2689-2739.
Aghion, Philippe, Patrick Bolton, and Steven Fries, 1999, Optimal Design of Bank Bailouts: The Case of
Transition Economies, Journal of Institutional and Theoretical Economics, pp. 51-70.
Benmelech, Efraim, Ralf Meisenzahl, and Rodney Ramcharan, 2017, The Real Effects of Liquidity During the
Financial Crisis: Evidence from Automobiles, The Quarterly Journal of Economics, 132(1), pp. 317–365.
Benmelech, Efraim, Carola Frydman, and Dimitris Papanikolaou, 2019, Financial frictions and employment
during the Great Depression, Journal of Financial Economics, 133, pp. 541–563.
Berger, Allen, Christa Bouwman, Thomas Kick, and Klaus Schaeck, 2016, Bank liquidity creation following
regulatory interventions and capital support, Journal of Financial Intermediation, 26, pp. 115-141.
Berger, Allen, and Raluca Roman, 2017, Did Saving Wall Street Really Save Main Street? The Real Effects
of TARP on Local Economic Conditions, Journal of Financial and Quantitative Analysis, 52, pp. 1827-1867.
Berger, Allen, and Raluca Roman, 2015, Did TARP banks get competitive advantages? Journal of Financial
and Quantitative Analysis, 50, pp. 1199-1236.
Berger, Allen, Tanakorn Makaew, and Raluca Roman, 2019, Do Business Borrowers Benefit from Bank
Bailouts? The Effects of TARP on Loan Contract Terms, Financial Management, 48(2), pp. 575-639.
Bernanke, Ben, 1983, Nonmonetary Effects of the Financial Crisis in the Propagation of the Great Depression,
The American Economic Review, 73(3), pp. 257-276.
Bordo, Michael, and Anna Schwartz, 2000, Measuring real economic effects of bailouts: historical
perspectives on how countries in financial distress have fared with and without bailouts, Carnegie-Rochester
Conference Series on Public Policy, 53(1), pp. 81-167.
Butkiewicz, James, 1995, The Impact of a Lender of Last Resort during the Great Depression: The Case of the
Reconstruction Finance Corporation, Explorations in Economic History 32(2), pp. 197-216.
Butkiewicz, James, 1997, The Reconstruction Finance Corporation, the Gold Standard, and the Banking Panic
of 1933, Southern Economic Journal, 66(2), pp. 271-293.
Callaway, Brantly, and Pedro Sant’Anna, 2021, Difference-in-Differences with multiple time periods, Journal
of Econometrics, forthcoming.
Calomiris, Charles, and Joseph Mason, 1997, Contagion and Bank Failures during the Great Depression: The
June 1932 Chicago Banking Panic, American Economic Review, 87(5), pp. 863-883.
Calomiris, Charles, and Joseph Mason, 2003, Consequences of Bank Distress During the Great Depression,
American Economic Review, 93(3), pp. 937-947.
Calomiris, Charles, Joseph Mason, Marc Weidenmier, and Katherine Bobroff, 2013, The effects of
reconstruction finance corporation assistance on Michigan’s banks’ survival in the 1930s, Explorations in
Economic History, 50(4), pp. 526-547.
Cirera, Xavier, Marcio Cruz, Elwyn Davies, Arti Grover, Leonardo Iacovone, Jose Ernesto Lopez Cordova,
Denis Medvedev, Franklin Okechukwu Maduko, Gaurav Nayyar, Santiago Reyes Ortega, and Jesica Torres,
Policies to Support Businesses through the COVID-19 Shock: A Firm Level Perspective, The World Bank
Research Observer, 36(1), 41-66.

38
Cole, Harold, and Lee Ohanian, 2004, New Deal Policies and the Persistence of the Great Depression: A
General Equilibrium Analysis, Journal of Political Economy, 112(4), pp. 779-816.
Cortes, Gustavo, and Weidenmeir, Marc, 2019, Stock Volatility and the Great Depression, Review of Financial
Studies, 32(9), pp. 3544-3570
Cotter, Christopher, 2021, Off the Rails: The Real Effects of Railroad Bond Defaults following the Panic of
1873, American Economic Association, Papers and Proceedings 111, pp. 508-13.
Dewatripont, Mathias, and Jean Tirole, 2012, Macroeconomic Shocks and Banking Regulation, Journal of
Money, Credit and Banking, 44(2), pp. 237-254.
Diamond, Douglas and Dybvig, Phillip, 1983, Bank Runs, Deposit Insurance, and Liquidity, Journal of
Political Economy, 91, 401–419.
Diamond, Douglas, and Raghuram Rajan, 2004, Bank Bailouts and Aggregate Liquidity, American Economic
Review, 92(2), pp. 38-41.
Dougall, Herbert, and Loring Farwell, 1940, A Review of Railroad Financing, 1920-1938: I. Volume and Cost
of Railroad Capital, The Journal of Land & Public Utility Economics, 16(2), pp.207-213.
Duchin, Ran, and Denis Sosyura, 2012, The Politics of Government Investment, Journal of Financial
Economics, 106, pp. 24-48.
Elenev, Vadim, Tim Landvoigt, and Stijn Van Nieuwerburgh, 2021, Can the Covid Bailouts Save the
Economy? Working Paper.
Ennis, H. and Keister, T., 2009, Bank Runs and Institutions: The Perils of Intervention, American Economic
Review, 99, 1588–1607.
Faccio, Mara, Ronald Masulis and John McConnell, Political Connections and Corporate Bailouts, Journal of
Finance, 61(6), pp. 2597-2635.
Fishback, Price, Shawn Kantor, and John Wallis, 2003, Can the Three R’s be Rehabilitated? A Comprehensive
Analysis, Program by Program, County by County, Explorations in Economic History, 40(3), pp. 278–307.
Fishback, Price, 2017, How Successful Was the New Deal? The Microeconomic Impact of New Deal Spending
and Lending Policies in the 1930s, Journal of Economic Literature, 55(4), pp. 1435-85.
Flood, Mark, 1998, United States Historical Data on Bank Market Structure, 1896-1955. Inter-university
Consortium for Political and Social Research, Montreal, Quebec.
Friedman, Milton, and Anna Schwartz, 1963, A Monetary History of the United States, 1867-1960. Princeton:
Princeton University Press.
Giesecke, Kay, Francis Longstaff, Stephen Schaefer, Ilya Strebulaev, 2011, Corporate bond default risk: A
150-year perspective, Journal of Financial Economics, 102, pp. 233-250.
Goolsbee, Austan and Alan Krueger, 2015, A Retrospective Look at Rescuing and Restructuring General
Motors and Chrysler, Journal of Economic Perspectives, 29(2), pp. 3-24.
Gorton, Gary, and Lixin Huang, 2004, Liquidity, Efficiency, and Bank Bailouts, American Economic Review,
94(3), pp. 455-483.
Granja, João, Christos Makridis, Constantine Yannelis, and Eric Zwick, Did the Paycheck Protection Program
hit the target? NBER Working Paper 27095.
Grossman, Emiliano, and Cornelia Woll, 2014, Saving the Banks: The Political Economy of Bailouts,
Comparative Political Studies, 47, 574-600.

39
Hellmann, Thomas, Kevin Murdock and Joseph Stiglitz, 2000, Liberalization, Moral Hazard in Banking, and
Prudential Regulation: Are Capital Requirements Enough? American Economic Review, 90(1), pp. 147-165.
Jones, Jesse, 1951. Fifty Billion Dollars: My Thirteen Years with the RFC (1932-45). Macmillan Company,
New York.
Kaplan, Edward, and Paul Meier, 1958, Nonparametric estimation from incomplete observations, Journal of
the American Statistical Association, 53 (282), pp. 457–481.
Kelly, Bryan, Hanno Lustig, and Stijn van Nieuwerburgh, 2016, Too-Systemic-to-Fail: What Option Markets
Imply about Sector-Wide Government Guarantees, American Economic Review, 106(6), pp. 1278-1319.
Lucas, Deborah, 2019, Measuring the Cost of Bailouts, Annual Review of Financial Economics, 11, pp. 85-
108.
Mason, Joseph, 2001, Do Lender of Last Resort Policies Matter? The Effects of Reconstruction Finance
Corporation Assistance to Banks During the Great Depression, Journal of Financial Services Research, 20(1),
pp. 77-95.
Mason, Joseph, 2003, The political economy of Reconstruction Finance Corporation assistance during the
Great Depression, Explorations in Economic History, 40, pp. 101-121.
Mason, Joseph and Daniel Schiffman, 2004, Too Big to Fail, Government Bailouts, and Managerial Incentives:
The Case of the Reconstruction Finance Corporation Assistance to the Railroad Industry during the Great
Depression in Too Big to Fail, Benton Gup (editor), Praeger Publishers, Westport, Connecticut.
Ohanian, Lee, 2009, What – or who – started the great depression? Journal of Economic Theory, 144(6), pp.
2310-2335.
Reconstruction Finance Corporation, 1959, Final Report on the Reconstruction Finance Corporation, U.S.
Government Printing Office, Washington.
Sant’Anna, Pedro, and Zhao, Jun 2020, Doubly robust difference-in-differences estimators, Journal of
Econometrics, 219(1), pp. 101-122.
Schiffman, Daniel, 2003, Shattered Rails, Ruined Credit: Financial Fragility and Railroad Operations in the
Great Depression, Journal of Economic History, 63(3), pp. 802-25.
Wallis, John, 1987, Employment, Politics, and Economic Recovery During the Great Depression, Review of
Economics and Statistics 69(3), pp. 516–20.
Wright, Gavin, 1974, The Political Economy of New Deal Spending: An Econometric Analysis, The Review
of Economics and Statistics, 56(1), pp. 30-38

40
Appendix
Table A.1 Cross-sectional regression
We regress the change in profitability, leverage, and (log) employment between 1929 and 1939 on characteristics fixed
in 1929. We include a dummy that yields one if the railroad has received (at least) one bailout between 1932 and 1939.
We add the average annual number of connections between 1932 and 1939. p-values are adjusted for heteroskedasticity
and clustered at the railroad-level. *, **, and *** denote significance at the 10%, 5% and 1% levels, respectively.
Profitability
Leverage
Employment
Wage Bill
(1)
(2)
(3)
(4)
Approval1932 – 1939
0.007
(0.326)
-0.045
(0.292)
0.025
(0.753)
0.006
(0.605)
Average Connections1932 - 1939
0.010
(0.124)
0.002
(0.962)
0.031
(0.646)
0.017*
(0.078)
Log(Total Assets)
-0.006
(0.282)
-0.028
(0.414)
0.109*
(0.085)
0.061***
(0.000)
Net Income / T.A.
-0.905***
(0.000)
0.705
(0.294)
1.277
(0.307)
-0.249
(0.162)
Leverage
-0.014
(0.483)
-0.202*
(0.078)
0.039
(0.852)
-0.002
(0.954)
Cash
0.367**
(0.031)
-2.119**
(0.031)
-1.328
(0.461)
-0.220
(0.388)
Log (Age, years)
0.001
(0.729)
0.014
(0.551)
-0.051
(0.242)
0.001
(0.817)
Volatility
0.018
(0.677)
0.204
(0.403)
-0.160
(0.724)
-0.073
(0.256)
Passenger / Total Revenue
-0.040
(0.198)
0.067
(0.706)
-0.091
(0.785)
-0.036
(0.440)
Employment
0.002
(0.716)
0.020
(0.576)
-0.117*
(0.084)
-0.067***
(0.000)
BondsDue1930 – 1934
0.042
(0.292)
0.093
(0.682)
-0.078
(0.854)
0.006
(0.915)
Region FE
Yes
Yes
Yes
Yes
Freight Composition
Yes
Yes
Yes
Yes
R squared
0.429
0.111
-0.061
0.352
Observations
115
115
115
115