Access to Collateral and The Democratization of Credit:

France’s Reform of The Napoleonic Security Code*

Kevin Aretz Murillo Campello∗∗ Maria-Teresa Marchica

Manchester Business School Cornell University & NBER Manchester Business School
kevin.aretz@mbs.ac.uk campello@cornell.edu maria.marchica@mbs.ac.uk

This Draft: May 19, 2017

Abstract

France’s Ordonnance 2006-346 repudiated the notion of possessory ownership in the Napoleonic Code,
easing the pledge of physical assets in a country where credit was highly concentrated. The reform was
undermined by “non-codified” laws that already allowed firms in large cities to pledge liquid assets
to factoring companies. Firms operating hard assets and with limited access to factoring significantly
increased their leverage following the reform, with the fraction of “zero-leverage” firms dropping from
90% to 30%. Small, profitable, and start-up firms benefitted the most. Credit market entrants observed
increases in performance, labor demand, and survival rates. Spatial point analysis shows that the
reform reached firms in rural areas, reducing capital access inequality across France’s countryside.

Key words: Security Laws, Contractibility, Collateral, Capital Structure, Bank Loans.
JEL classification: G32, K22, O16.

*We thank Manuel Adelino, Nittai Bergmann, Charles Calomiris, Gilles Chemla, Alberta Di Giuli, Mara Faccio,
Edith Ginglinger, Gaurav Kankanhalli, Mauricio Larrain, Kai Li, Ulf Lilienfeld-Toal, Roni Michaely, Matteo
Millone, Abhiroop Mukherjee, Giovanna Nicodano, Steven Ongena, Maurizio Pisati, Zacharias Sautner, Reinhard
Schmidt, Antoinette Schoar, David Thesmar, Giuseppe Vittucci, and Baolian Wang for their constructive and
helpful insights. Comments from participants at the 2017 American Economic Association Meeting (Chicago),
the 2017 CEPR Spring Symposium (London), the 2016 Edinburgh Corporate Finance Conference, the 2015
European Banking Center Conference (Tilburg), the 2015 European Finance Association Meeting (Vienna),
the 2016 Financial Intermediation Research Society (FIRS) Conference (Lisbon), the 2015 International Con-
ference on Credit Risk Valuation (Venice), the 2016 UBC Summer Finance Conference (Vancouver), the 2016
Western Finance Association Meeting (Park City) as well as seminar participants at Birmingham University,
Bristol University, Columbia University, ESCP Paris, Free University of Bozen, IDC Herzliya, Paris-Dauphine
University, St. Andrews University, University of Piraeus, and WHU Koblenz are also acknowledged. We are
grateful to Marie-Elodie Ancel from Université Paris Est Créteil, Rod Cork from Allen & Overy (Paris), and
Philip Wood from Allen & Overy (London) for insightful discussions about French laws.
**Corresponding author: Cornell University, 114 East Avenue, 381 Sage Hall, Ithaca, NY 14853. Phone: 607-
255-1282. E-mail: campello@cornell.edu.
“To allow for a more equal access to credit, I wish to see a rapid easing of the pledge regime.
Do I need to remind you that currently a borrower, for example, a company, has to hand over
the property that it pledges? Do you know a modern country that works that way?”

— Pascal Clément, Minister of Justice, addressing the French Parliament on June 22, 2005.

1 Introduction
Economic elites are thought to exert pressure over the design of legal contracting frameworks,
seeking arrangements that benefit their interests (Rajan and Zingales (2003)). Manifestations
of discriminatory contracting schemes include usury laws and rules governing the validation of
collateral in credit transactions. To date, several countries use security laws derived from the
Napoleonic Code, a regime predicated on “possessory ownership.” Under the highly-formalized
Napoleonic Code, physical assets are deemed to be “unique,” “whole,” and “non-transferrable”
(see Omar (2007) and Ancel (2008)). These legal fictions limit the types of security interests
that can be written on assets, favoring large, well-established, well-connected incumbents over
small, young, innovative newcomers. The repeal of “endowed, perverse” contracting frameworks
is seen as a route to democratizing credit access, but ultimate economic consequences are hard
to gauge (Demirgüç-Kunt and Levine (2008) and Rajan (2009)). This paper shows how a recent
reform in France informs knowledge about the link between financial contracting, access to
credit (level and distribution effects), and real economic outcomes.
Ordonnance 2006-346 derogated the notion of possessory asset ownership in France, in
existence since 1804. By doing so, the 2006 reform allowed French firms to control and operate
(in-house) physical assets pledged to third parties. This seemingly simple statutory change sig-
nificantly enlarged the menu of assets that firms could pledge in credit transactions, particularly
hard movable assets used in modern business operations (e.g., machinery and equipment). In
addition to expanding the menu of assets that could be collateralized, the new regime allowed
for security interests to be charged to more than one party, making it feasible for loans to
be syndicated under multiple creditors, with multiple priority schemes, and multiple maturity
structures. In further allowing for rechargeable interests, the reform also enhanced the pledge-
ability of hard immovable assets (land and buildings). Notably, the reform did not introduce
changes in the balance of power between contracting parties, in asset seizure options, or in
judicial intervention. This differentiates it from most recent reforms, which have promoted the
notion of “strengthening creditors’ rights” as a means to enhance credit access.
Changing the legal framework governing asset ownership and alienability meant that firms
were discretely endowed with “new assets” that they could pledge as collateral in credit transac-
tions. Yet the new law did not affect all French firms equally. Long before Ordonnance 2006-346
was brought to the policy debate, firms located in France’s financial centers had successfully
lobbied for legal exceptions that allowed them to pledge certain types of liquid assets as collat-
eral. These exceptions were (purposely) not formally reconciled with the Napoleonic Code, and
labeled “non-codified” security laws. Chief among them were laws that allowed firms to pledge
cash, securities, and accounts receivables to factoring companies.1 Most of the collateral posted
by French firms before Ordonnance 2006-346 consisted of liquid assets under non-codified credit
agreements (see Davydenko and Franks (2008)).2
The wrinkles in the process through which France reformed its security code system allow for
unique insights into the distributional and wealth effects of easing access to collateral. We build
on the prior that firms whose operations relied most intensely on hard assets would be favored
by a reform that distinctly enhanced their ability to pledge that class of assets to creditors.
Naturally, firms did not need to use hard assets to secure credit; they could use liquid assets
instead. As such, a second layer of our strategy uses the observation that French firms needed
access to factoring services to pledge their liquid assets. Critically, the operations of factoring
companies were legally constrained to identifiable geographic areas and targeted businesses
locally.3 As a result, firms located in areas with no offerings of factoring services had a difficult
time pledging any of their assets — either liquid or hard — before Ordonnance 2006-346. In
this setting, we contrast firms along the types of assets more intensely used in their production
processes (hard versus liquid) and also according to the nature of the contracts that they sign (se-
cured versus unsecured, long- versus short-term), their legal status (private versus public), and
their geographic location (distance from factoring companies), among several dimensions that
help us identify how collateral contracting shapes credit taking and its economic consequences.
1
Two examples of such laws are the “Dailly Law” and the “Pledge of Ready Money,” put in place in the
1970s. We discuss these financing schemes shortly.
2
As discussed below, alternative financing arrangements, such as leasing, were not commonly used by French firms.
Practitioners attribute this to the “excessive protection” given to lessees regarding the termination of lease contracts.
3
To date, over 90% of the French factoring companies have only one physical branch. Only one factoring
company had more than a couple dozen branches in 2006, all confined to three large cities.

1
 We first study the effect of Ordonnance 2006-346 on firm-level (quantity) and contract-level
(price) credit data within a differences-testing framework. We find that the reform significantly
increased the debt-taking of French firms with no access to public security markets along both
the intensive margin (leverage ratios) and the extensive margin (propensity to take out any debt
at all). Notably, the reform only impacted long-term debt-taking (secured by hard assets). Using
firms’ fixed assets intensity to identify effect heterogeneity, we show that the increase in debt
ratios and the decline in the proportion of “zero-leverage” firms occurred primarily among
“high-fixed assets” firms. In particular, while the long-term leverage ratio of high-fixed assets
intensity firms rose 8 percentage points after the reform, the long-term leverage ratio of low-
fixed asset firms rose by a mere one point. The 7% difference is remarkable when compared to
the pre-reform average leverage ratio of only 2%. Confirming the logic of our test, short-term
leverage (not secured by hard assets) did not change across high- or low-fixed asset firms. Using
the geographic proximity to factoring companies as an additional identification wrinkle, we show
that the documented increase in debt ratios and the decline in the proportion of zero-leverage
firms occurred among firms located far away from factoring companies; that is, firms that had
the hardest time pledging any kind of assets before the collateral reform. Our results account
for firm as well as industry-year dynamic effects and are robust to a battery of checks.
We use loan-level data to study whether the reform led to contract term changes. Difference-
in-differences estimations show that secured loans (loans that were directly treated by Ordon-
nance 2006-346 ) became less expensive and were given longer maturities compared to unsecured
loans after the reform. The interest rate mark-ups of secured loans dropped by 146 basis points
relative to mark-ups of unsecured loans (the pre-reform mean was 213 basis points). Their ma-
turities doubled. Consistent with the idea that asset-backed contracting became more flexible,
the number of lenders involved in secured transactions increased relative to those in unsecured
transactions. Our results imply that banks became willing to extend secured loans at more
favorable terms after the 2006 collateral reform, speaking to the argument that an increase in
the collateral-offer space has an easing effect on multiple dimensions of credit access.
The next step in our analysis is to study whether the reform-led access reached firms previ-
ously rationed in the credit market (“credit democratization”). We perform several tests on this
front. We first look at the demographics of firms that entered the capital market with the reform.

2
In particular, we contrast firms that never borrowed before the reform and borrowed in every
year after the reform (“new borrowers”) with other firms that operate in their industries. New
borrowers were smaller, more cash-constrained, less risky, and more profitable firms, located
farther from large cities. Our analysis suggests that the collateral reform benefitted seemingly
creditworthy firms that were previously cut-off from the credit market. We then test whether the
reform reached a class of firms that is particularly subjected to credit rationing: start-up firms.
Our tests show that the average long-term leverage of start-ups incorporated after 2006 is 1.4 per-
centage points higher than that of start-ups incorporated before 2006, when the sample average
was 2.6%. Also, the proportion of start-ups with no long-term debt in their year of incorpora-
tion drastically dropped following the reform. Our results suggest that Ordonnance 2006-346
allowed new entrepreneurs to raise funds in the credit market upon entering their industries.4
Our main credit democratization test is a geography-based analysis of the reform’s effects.
Several studies emphasize the importance of firm location for financial contracting. A higher
density of financial institutions, in particular in and around urban centers, lowers the informa-
tional costs of credit acquisition (Garmaise and Moskowitz (2004)). As such, firms located in
rural areas face a significant disadvantage in accessing credit. We use French department-level
data to trace the effects of Ordonnance 2006-346 on credit access. Spatial point analyses show
that firms located farther from the main French metropolitan areas benefitted the most from
the collateral reform. We also compute a Gini index of corporate credit access to gauge whether
the reform reduced inequality in credit access within and across departments. Before the reform,
91 out of 96 French departments had a Gini index exceeding 0.90, implying that the credit
market in virtually every department was dominated by a handful of large firms.5 After the
reform, only 19 departments registered an index above 0.90. Notably, the largest drops occurred
in rural departments, far from the main French urban areas and financial centers.
A key question is whether the increase in credit access had tangible economic consequences.
To tackle this issue, the final part of our analysis tracks the real effects of the reform on various
fronts. At the microeconomic level, we find that high-fixed assets firms increased their invest-
4
Our evidence complements findings by Schmalz et al. (2017) showing that increases in home collateral
values lead to increases in the probability of becoming an entrepreneur in France.
5
France’s Gini index of corporate credit inequality was 0.95 before 2006, a high figure when compared to
0.82 in other European Civil Law countries (Belgium, Italy, Portugal, and Spain).

3
ment significantly more than low-fixed assets firms after Ordonnance 2006-346. They also hired
more workers, particularly in non-urban areas of the country, with average wages increasing in
those areas. Our evidence also suggests that increases in capital and labor enabled high-fixed
assets firms to pursue more profitable, less risky investments. This is an important result from
an economic policy perspective as it implies that the “new, marginal” investments made pos-
sible by the reform ultimately contributed to the increase in corporate survival rates that we
also document in the data. Indeed, while high-fixed assets firms were more likely to fail than
low-fixed assets firms before the reform, that relation is significantly reversed afterwards.6
Tracing effects of the reform at the macro level, we run two sets of tests. First, following Wur-
gler (2000), we use the sensitivity of capital investment growth to value added growth as metric
of the efficiency of the cross-sectoral capital allocation process. Wurgler’s efficiency sensitivity
metric jumps from 0.48 to a post-reform level of 0.85. We also use the external financing depen-
dence proxy of Rajan and Zingales (1998) as an additional capital allocation efficiency metric.7
We find that the reform raised long-term borrowing more pronouncedly in sectors that were
more highly dependent on external financing. The bulk of our evidence suggests that the reform-
induced changes in the collateral regime improved the capital allocation process in France.
Our study contributes to the Law and Finance literature on various fronts. Most existing
studies analyze cross-country heterogeneity in bankruptcy regimes (LaPorta et al. (1998) and
Djankov et al. (2007)) or within-country changes to bankruptcy regimes (Lilienfeld-Toal et
al. (2012) and Vig (2013)). Our focus, instead, is on collateral, which carries different policy
implications. We show how collateral regime changes may spur the democratization of credit,
allowing small, profitable, start-up firms, and firms located in the countryside to access new
credit facilities. We also show how collateral regime changes may bring about positive real-side
implications at both micro and macroeconomic levels. Critically, these outcomes can be achieved
dispensing with the need to introduce — often controversial and costly — changes in the legal
powers of contracting parties or changes in the rule governing judicial intervention.
Our study also adds to recent research looking at how collateral affects bank credit. Cerqueiro
6
Our findings are in contrast with work showing how credit expansions tend to reach borrowers with worse
economic prospects, generating waves of default in the future (e.g., Mian and Sufi (2009)).
7
This proxy captures an industry’s demand for external funding. The working assumption is that when the
economic system operates efficiently, capital will be directed toward sectors that need more external funding.

4
et al. (2016) show that reductions in collateral values trigger increases in interest rates and reduc-
tions in credit limits. Calomiris et al. (2017) show that bank lending secured with movable assets
is lower in countries with weaker collateral protection. Our study is different in that we focus on
firms (rather than banks), allowing us to make inferences about real-side implications of collater-
alized debt-taking. We are able to show, for example, that a collateral-menu change that makes
it easier to pledge hard assets led firms whose asset composition is skewed towards those assets to
borrow more and invest into more profitable, less risky projects. Closer to our paper, Campello
and Larrain (2016) look at a collateral regime change in an emerging market, Romania. In
contrast to their paper, our setting and data allow us to study the effects of collateral on several
key dimensions, such as loan contract terms, start-ups’ access to credit, labor market outcomes,
firm survival, and aggregate capital allocation. The spatial analyses we employ further provide
unique insights into credit access distribution and democratization across an entire country.
Our paper is also related to the literature linking credit and economic development. Some
studies show that credit expansion may trigger waves of defaults and repossessions (White
(2007), Mian and Sufi (2009), Keys et al. (2010), and Assunção et al. (2014)). Others show that
credit easing may lead to more business formation and higher economic activity (Benmelech and
Moskowitz (2010) and Chatterji and Seamans (2012)). We add to this literature by showing how
collateral-induced credit creation may shape the capital allocation process: not only is capital
directed toward growth industries, but also toward industries starving for external funding.
The balance of our article is organized as follows. Section 2 discusses the French collateral
reform. Section 3 describes our data and empirical methodology. Section 4 analyzes the effect
of the reform on credit access. Section 5 analyses the demographics of new borrowers. Section 6
performs a geographic analysis of the reform’s effects. Section 7 discusses real-side implications.
Section 8 contains robustness checks. Section 9 concludes.

2 Institutional Setting: Security Laws in France

2.1 The Napoleonic Code of 1804

Until recently, security interests in France were governed by the 1804 Napoleonic Code. The Code
recognized two forms of security interests, the “hypothec” and the “charge.” A hypothec could

5
only be taken over real estate property, while a charge could be taken over all types of assets.8 In
spirit and practice, the early 19th century security code focused on interests over land. At that
time, the French economy was centered around land-based activities and lawmakers deemed
interests over other assets as “of little relevance” (see Omar (2007)). As the need for modern
financing demanded new securitization techniques, the legislature filled gaps in the law with ad
hoc regulation (Haimo (1983)). The cumulative effect of this approach was a lack of consistency
in requirements for creating and perfecting security interests over different types of assets.
The Napoleonic Code was predicated on the notion that borrowers had to be protected from
exploitative creditors, while creditors had to be protected from borrowers refusing the surren-
der of collateral in default. In reconciling such interests, regulators established highly formal,
costly procedures for the creation of security rights. Among other idiosyncracies, the 1804 Code
would not recognize non-possessory charges since assets pledged without dispossession could
lead future creditors to “misapprehend a borrower’s creditworthiness” (see Haimo (1983)).9
Critically, ruling out non-possessory charges meant that borrowers had to physically transfer
pledged assets for creditors to validate — in the legal jargon, “perfect”— a security interest.
This statute of the law had profound implications for borrowing. In particular, it implied that
firms could not pledge assets to third parties while at the same time continuing to freely use
and dispose of those assets in their regular (in-house) operations. Moreover, because assets were
deemed as “unique” and “non-substitutable,” firms could not offer similar assets (nor even cash
equivalents) in securing a credit transaction. Security interests over “fungible assets” (such
as oil, steel, etc.) or “future assets” (such as machinery and equipment under construction, or
inventory in process) were also disallowed because such assets could not be easily identified
and physically transferred to creditors. Finally, the Code would not recognize rechargeable
hypothecs since these could lead to conflicts about priority (see Ancel (2008)).
Alternative sources of financing not requiring collateral, such as leasing, had been of rather
limited use in France. This has been attributed to the strong protection afforded to the lessees
regarding the termination and renewal of lease agreements. At the expiration of a commercial
lease, for example, the lessee is entitled to an additional term of 9 years provided the lessee has
8
See Boughida et al. (2011) for more details on French security interests.
9
The refusal to recognize non-possessory charges is grounded in the notion that only the physical possession
of production assets would unequivocally demonstrate a firm’s creditworthiness.

6
occupied the premises for 3 years; else, lessors pay a fine. In the U.S. and German systems, in
contrast, lessees are only entitled to a new lease at expiration when that option was part of the
original agreement; lessors are not obliged to pay any compensation (Lofstedt and Baum (1993)).
The average proportion of France’s gross fixed investment funded through leasing over our
sample period was about 11%, both before and after the reform. This figure is remarkably low
when compared to 26% for the U.S., 17% for Germany, and 13% for other European Civil Law
countries over the same period (cf., Leaseurope (2009) and White Clark Group (2015)).10

2.2 Reforming the Napoleonic Code: Ordonnance 2006-346

France’s outdated security laws became a topic of much debate in 2005 (see Renaudin (2013)).
In March 2005, President Jacques Chirac asked the Minister of Justice to reform the country’s
security regime. In April, an official working group of academics and practitioners submitted
recommendations on how to best achieve that goal (the Grimaldi Report). In June, the French
Parliament passed Law 2005-842 on “Trust and Modernization of the Economy,” enabling the
government to reform security laws by way of Ordonnance (government order). Under this
exceptional mandate, the central government could unilaterally pass laws on economic matters
without congressional approval. In March 2006, the French government enacted a sweeping
reform of the country’s security laws.
Ordonnance 2006-346 transformed the law and practice of secured credit transactions in
France. The reform repudiated the notion of possessory ownership, allowing for parties to write
contracts contemplating non-possessory charges and security interests over various types of
assets. The new statute also allowed security interests to be rechargeable to the same or new
creditors, either simultaneously or consecutively. Importantly, the new law did not introduce
“floating charges” (charges applying to the “general collection” of corporate assets). As such,
the formal identification of assets legally recognizable and eligible for collateralized charges
remained a key issue for credit access (see Herbet and Sabbah (2006) and Ancel (2008)).
The 2006 reform had originally also aimed at strengthening creditors’ rights by lifting the
ban on out-of-court seizures of collateral in default. However, the French Parliament appointed
10
Interestingly, France is home to some of the largest leasing companies in Europe (Société Générale Leasing
and BNP Paribas Leasing). Their businesses, nonetheless, take place primarily outside of France.

7
two separate groups — accountable to different branches of government — to modernize the
security laws and the bankruptcy laws, and no effort was made to ensure that reforms in the two
areas would be synchronized. As such, several years after the 2006 reform, the bankruptcy code
still suspended a creditor’s right to seize collateral out-of-court once a firm files for bankruptcy.
The failure to strengthen creditors’ rights in default makes the French reform unique: a reform
that enlarged the contracting space without altering the balance of power between contracting
parties, seizure procedures, or court involvement.11

2.3 Undermining the 2006 Reform: Non-Codified Security Laws

Managers of large companies in major French cities had long recognized that the 1804 Security
Code was outdated and costly to work with. In the 1970s and 1980s, their concerted lobbying
spurred security laws that were purposefully not added to the Napoleonic Code.12 The most
popular of these “non-codified” laws were the Dailly Law and the Pledge of Ready Money. The
Dailly Law (named after the Parisian businessman Etienne Dailly) allowed firms to assign their
receivables to non-banking financial institutions in securitizing debt contracts. The Pledge of
Ready Money allowed them to do the same using cash and cash equivalents. Both laws were
remarkably modern insofar as permitting borrowers to take out security interests over fungible
and future assets and granting super-priority rights to secured creditors in bankruptcy. Their
major limitation was that their use was restricted to liquid assets.
In practice, non-codified credit schemes allowed firms to sell receivables and other liquid
assets to factoring companies.13 In a standard factoring agreement, a factor enters an exclusive
contract with a firm, obliging the firm to pass on all buy orders received by it to the factor.
The factor decides whether the firm is allowed to accept an order. If the firm accepts an order,
the factor purchases the associated receivables at a discount, taking on the responsibility of
11
For completeness, we did an extensive search on laws and regulations dealing with contracting in France
and identified another noteworthy proposal: the “2005 Safeguard Provision.” The 2005 rule gave creditors
more influence over the bankruptcy process by introducing creditor committees in firms with more than 150
employees. Researchers have argued that the 2005 regulation failed to achieve its stated goals (see, e.g., Plantin
et al. (2013)). Regardless, rerunning our empirical tests on the subsample of French firms unaffected by the
regulation (those with less than 150 employees; 80% of our sample) does not change our inferences in any way.
12
Bypassing the formal Code was a way to facilitate their implementation on an “exceptional basis.”
13
A Dailly assignment of receivables could only be granted by a borrower (and not by a guarantor or a third
party security grantor) and only in favor of a French licensed factoring credit institution.

8
Figure 1. Comparison of French Security Law Regimes The table compares the old French security law
introduced by the Napoleonic Civil Law Code in 1804, the non-codified laws enacted in the late 1970s and early
1980s, and the new security law introduced by Ordonnance 2006-346 in 2006.

administering and monitoring them. If the buyer defaults, the factor bears the loss, since it is
not entitled to seek compensation from the firm. The nature of factoring agreements requires
that factors must have accurate information about a firm and its customers on a regular basis
(Mian and Smith (1992) and Smith and Schnucker (1994)).
A factor’s need for accurate and timely information poses a problem for factoring in France
since the operations of factoring companies were geographically fragmented; a condition stem-
ming from the strict regulation of the country’s financial system. Of the 96 French departments,
only four had more than a dozen factors in operation in 2006. Those four departments included
France’s largest financial centers: two in the greater Parisian area, Lyon, and Lille. The vast
majority of factoring firms had only one branch. These conditions limited the geography of
non-codified credit contracting since factors had a hard time obtaining accurate and timely infor-
mation about firms located in France’s countryside. For practical purposes, France’s non-codified
laws only facilitated credit access to firms located in a limited number of metropolitan areas.

2.4 Identification Strategy

Figure 1 provides an overview of the various security law regimes in place in France since
1804. Relative to the Napoleonic Code, the non-codified laws of the 1970s and 1980s enhanced
the pledgeability of liquid assets by allowing firms to create security interests over accounts
receivables and other cash-like assets. As discussed, they contemplated firms that were physically
close to factoring companies. The 2006 reform enhanced the pledgeability of hard assets. It
did so most notably by allowing for the creation of non-possessory security interests over hard
movable assets, such as equipment and machines. It also allowed for rechargeable security
interests over hard immovable assets, such as land and buildings.

9
 We use the various wrinkles in the French security system to design two firm-level identi-
fication strategies. The first uses the observation that, while hard assets could not be easily
pledged as collateral before the 2006 reform, liquid assets could already be pledged. Accordingly,
the 2006 reform is expected to have affected more significantly those firms whose pre-reform
operations employed more fixed assets, relative to firms operating more liquid assets. The
second strategy uses the observation that firms needed the services of a factoring company to
pledge their liquid assets. The combination of these two conditions implies that Ordonnance
2006-346 should have affected more significantly hard-assets intensive firms located far away
from factoring companies, as these firms had a hard time pledging either hard or liquid assets
prior to the reform. Differently put, we use the geographic proximity to factoring companies
as a factor modulating the differential effect of the collateral reform on firms with different
asset compositions. As an added contrast, we use a contract-level identification strategy that
compares loans that were directly treated by the collateral reform mandate (secured loans)
with loans that were not (unsecured loans).

3 Data and Methodology

3.1 Data

Our firm financial data come from Bureau van Dyck’s AMADEUS database, which contains
comprehensive information on public and private firms from 35 European countries.14 We note
that each release of the AMADEUS database only covers the most recent eight years of financial
data per firm and does not contain data on firms that became inactive more than four years
before. To avoid survivorship biases, we retrieve all historical data from each April release of
the database between 2001 and 2011. We pool these data, deleting duplicate entries. We collect
data on French firms as well as on firms in other European Civil Law countries (Belgium,
Italy, Spain, and Portugal). Our baseline tests exclude public firms as these rarely borrow
under secured debt facilities.15 We exclude firms in the financial, services, utilities, and public
14
We have access to annual releases of AMADEUS Top 250,000. For France, the database includes all
companies that meet at least one of the following criteria: (1) revenues of at least e15 million, (2) total assets
of at least e30 million, and (3) at least 200 employees. Once the company enters the database, AMADEUS
backfills its accounting information for the preceding years.
15
We use public French firm data and private non-French firm data to conduct placebo tests described later.

10
administration sectors. Our base sample comprises 34,910 unique firms over the 2001–2009
period. To run loan-level tests, we merge AMADEUS with the LPC-Dealscan database.16 We
collect data on the addresses of French firms from AMADEUS and BANKSCOPE. We use the
United Nations’ General Industrial Statistics (UNIDO) database (INDSTAT-4) to obtain data
on capital formation and value added for 124 four digit-ISIC manufacturing industries.

3.2 Variable Construction

3.2.1 Firm-Level Variables

We use total, long-term, and short-term leverage to study the debt financing effects of the
reform. TotalLeverage is the sum of short-term debt (mnemonic loan) and long-term debt
(ltdb) divided by total assets (toas). ShortTermLeverage (LongTermLeverage) is defined as
short-term (long-term) debt divided by total assets. We further define two dummy variables,
DummyNoShortTermDebt and DummyNoLongTermDebt, which equal one if the corresponding
debt variable is zero. FixedAssets is fixed assets (fias) divided by total assets.
To gauge firm-level real effects, we consider firms’ investment, employment, sales, profitability,
and risk. Investment is the change in tangible fixed assets (tfas) from the prior fiscal year end
to the current fiscal year end plus depreciation (depr), scaled by the average of total assets
taken over the two fiscal year ends. Employees is the log number of employees (emp). Sales is
the log of sales (sale). Profitability is earnings before interest and taxes (ebta) divided by total
assets. ProfitVolatility is the standard deviation of Profitability over the previous four years. In
addition to Employees, we consider two other labor-related variables. AverageWage is the log of
total wages (staf) scaled by the number of employees. FactorProductivity is the residual from a
Cobb-Douglas production function.17 Our models use standard control variables. Among others,
Size is the log of total assets, and Age is the log of the difference between the current year and
the year of incorporation. Continuous variables are winsorized at the first and last percentiles.
16
We consider revolver and term loans. A term loan is a loan that has to be repaid according to a fixed
repayment schedule. A revolver allows for a specific amount to be withdrawn, repaid, and redrawn again in any
manner or number of times.
17
In the Cobb-Douglas production function, output is sales, capital is fixed assets, and labor is the number
of employees. Factor elasticities vary across sectors. Each sector’s labor elasticity is measured as the ratio of
labor compensation to value added, while each sector’s capital elasticity is one minus its labor elasticity.

11
 Our analysis also uses the availability of factoring companies per French postal code. We
use resources from geonames.org to identify the longitude and latitude associated with each
French postal code (4,445 postal codes). NoFactoringAvailable is a dummy variable equal to
one if there are no factoring companies in the postal code area in which a firm is registered, else
zero. To derive a firm’s distance to the closest of the ten largest French cities (Dist10Cities),
we calculate the spherical distance between points i and j using:18

DISTi,j = arccos(deglatlon) × r, (1)

where deglatlon is given by: cos(LATi )cos(LGTi )cos(LATj )cos(LGTj ) + sin(LGTi )sin(LGTj )
+ cos(LATi )sin(LGTi )cos(LATj )sin(LGTj ), LGTx and LATx are the longitude and latitude
of point x, respectively, and r is the radius of the earth (see Coval and Moskowitz (2001)).

3.2.2 Contract-Level Variables

Our contract-level analysis uses loan mark-ups, loan maturity, and the number of lenders in a
syndicate as outcome variables. LoanSpread is the log of the sum of a loan’s coupon and annual
fees scaled by its nominal value minus the six-month LIBOR rate, in basis points. LoanMaturity
is the log of the difference between the loan’s maturity date and its initiation date, in months.
NumberLenders is the log of the number of lenders involved in a loan agreement. Secured is
a dummy variable that equals one for secured loans (loans treated by Ordonnance 2006-346 ).
As controls, we include firm size, age, profitability, total leverage, the existence of a credit
rating, loan size, and loan type. Rating is a dummy variable that equals one if the firm is rated
(indicating a higher credit quality). LoanAmount is the log of the notional loan value. LoanType
is a dummy variable equal to one for term loans and equal to zero for revolver loans.

3.2.3 Industry-Level Variables

In studying capital allocation efficiency, we first use four-digit ISIC industry-level data for
French industrial sectors. InvestmentGrowth (ValueAddedGrowth) is the natural log of the ratio
18
France’s ten largest cities are: Paris, Lyon, Marseille, Toulouse, Lille, Bordeaux, Nice, Nantes, Strasbourg,
and Grenoble.

12
of current gross fixed capital formation (value added) to its one year-lagged value.19 Second, we
create an industry-level external financial dependence index similar to Rajan and Zingales (1998).
The index is calculated as the ratio of capital expenditures net of cash flows to total capital
expenditures of the median U.S. firm within each three-digit SIC over the 1975–2005 period.

3.3 Methodology

We use differences estimations to gauge the effects of the collateral reform. In our first firm-level
identification strategy, we compare outcome variables of interest across high- versus low-fixed
assets utilization firms over the pre-reform (2001–2005) versus post-reform (2006–2009) periods.
We estimate a series of regressions including fixed effects, which can be written as:

Yi,t = αi + αk,t + βP ostt × T reatedi + Xi,t γ + εi,t , (2)

where Yi,t is the value of the outcome variable for firm i in year t, with Yi,t ∈ {LongTermLeverage,
ShortTermLeverage, DummyNoLongTermDebt, DummyNoShortTermDebt, Investment, Employ-
ees, Sales, Profitability, ProfitVolatility}. P ostt is a dummy variable that equals one for year 2006
onwards, and zero otherwise. Using our first identification strategy, we set T reatedi equal to
T reatedHighF
i
ixedAssets
, a dummy variable that equals one if firm i’s fixed assets-to-total assets ra-
tio is in the top quartile of the year 2005 distribution, else zero.20 Using our second identification
strategy, we set T reatedi equal to T reatedN
i
oF actoringAvailable
, a dummy variable that equals one if
firm i is registered in a geographic area (postal code) without factoring companies. X is a vector
of firm-specific controls. αi and αk,t are firm- and interacted year-industry-fixed effects, respec-
tively, where the industry effects are based on the NAF (Nomenclature des Activités Françaises)
code system. We cluster standard errors at the firm level. The β coefficient in Eq. (2) can be in-
terpreted as a regression-based DID estimate after accounting for controls and fixed effects.
We subsequently combine the above two identification strategies and run regressions that
19
Both variables are GDP-deflated. Gross fixed capital formation is defined as the cost of new and used fixed
assets minus the value of sales of used fixed assets, where fixed assets include land, buildings, and machinery and
equipment. Value added is the value of shipments of goods (output) minus the cost of intermediate goods and
required services, with adjustments made for inventories of finished goods, work-in-progress, and raw materials.
20
Using alternative quantiles (such as quintiles or terciles) does not alter our inferences. The choice for our base-
line identification variable is guided by the natural trade-off between bias and efficiency one faces in these cases.

13
interact both treatment indicators. These models can be written as:

Yi,t = αi + αk,t + βP ostt × T reatedHighF

i
ixedAssets
+ δP ostt × T reatedN
i
oF actoringAvailable

+γP ostt × T reatedHighF

i
ixedAssets
× T reatedN
i
oF actoringAvailable
+ Xi,t η + εi,t . (3)

with γ returning the DIDID estimate of interest.

We use a model analogous to Eq. (2) to perform our loan-level analysis:

Zi,l,t = αk,t + βP ostt × T reatedSecured

i,l,t + Xi,t δ + Wi,l,t γ + εi,l,t , (4)

where Zi,l,t is the value of the outcome variable for loan l taken out by firm i in year t, with
Zi,l,t ∈ {LoanSpread, LoanMaturity, NumberLenders}. T reatedSecured
i,l,t is a dummy variable that
equals one for secured loans and zero for unsecured loans. In addition to firm-specific controls,
this model also includes loan-contract controls, contained in the vector Wi,l,t .21

4 Financial Outcomes
4.1 Debt Usage

4.1.1 Descriptive Statistics

Figure 2 depicts the mean total leverage ratios of firms in different fixed assets quartiles over
the sample period. The figure suggests that the security reform triggered an upward jump in
debt-taking starting in year-end 2006 (the reform year). In particular, the total debt-to-total
assets ratio of our sample firms increased from a pre-reform average of 9% to a post-reform
average of 12%. Remarkably, the effect of the reform is monotonically increasing in the rate
with which firms employ fixed assets in their production process: the higher the use of fixed
assets, the larger the debt-taking after the reform.
We next split total leverage into short-term and long-term leverage. Short-term debt in
France is secured by liquid — rather than hard — assets (see Vernimmen et al. (2015)) and
is thus not expected to be directly affected by the reform. Accordingly, Figure 3 shows that
the mean short-term leverage ratios of all fixed assets quartiles remain stable over our sample
21
Contract-level analyses like ours do not allow for firm-fixed effects because they are based on samples of
loan originations where most firms contribute only one observation (see, e.g., Ivashina and Scharfstein (2010)).

14
Figure 2. Evolution of Mean Total Leverage This figure shows the evolution of the mean total leverage
ratio by fixed asset quartile (Q) over the 2001–2009 period. Q1, Q2, Q3, and Q4 indicate the first, second, third,
and fourth fixed assets quartile, respectively.

Figure 3. Evolution of Mean Short-term vs.Mean Long-term Leverage This figure shows the evolution
of mean short-term and mean long-term leverage by fixed assets quartile (Q) over the 2001–2009 period. Q1,
Q2, Q3, and Q4 indicate the first, second, third, and fourth fixed assets quartile, respectively.

period. Long-term leverage ratios, in contrast, jump upward in the reform year, from a pre-
reform average of 2% to a post-reform average of 7%. As in Figure 2, effect magnitudes increase
monotonically over the range of fixed assets quartiles. While the long-term leverage ratio of
the highest fixed assets quartile firms rises 8 percentage points from the pre-reform to the
post-reform period, that of the lowest fixed assets quartile firms increases a mere one point.
We also study the reform’s effect on corporate debt-taking along the extensive margin. To do
so, we consider the evolution of firms with no short-term or no long-term debt (“zero-leverage
firms”). Figure 4 shows that the proportion of zero short-term debt firms across all fixed-assets
quartiles remains low and stable throughout the sample period. In contrast, there are striking
reductions in the proportions of zero long-term leverage firms in the reform year. In particular,

15
Figure 4. Evolution of the Proportions of Zero Short-term Leverage vs. Zero Long-term Leverage
Firms This figure shows the evolution of the proportion of zero short-term and zero long-term leverage firms
by fixed assets quartile (Q) over the 2001–2009 period. Q1, Q2, Q3, and Q4 indicate the first, second, third,
and fourth fixed assets quartile, respectively.

the proportion of zero long-term leverage firms in our sample decreased from a pre-reform aver-
age of 89% to a post-reform average of 42%. The magnitude of those reductions also increases
monotonically over the range of fixed-assets quartiles, with the highest fixed assets quartile
firms observing the largest reduction (from 88% to 32%).

4.1.2 Regression Results

Table 1 shows the results of DID regressions that use fixed-assets intensity (columns (1) through
(4)) and physical proximity to factoring companies (columns (5) through (8)) as identification
wrinkles. Across all models the estimates in row (1) of Table 1 point to significant, positive
effects of the reform on long-term debt-taking — secured by hard assets — across high-fixed
assets firms and firms located in areas where there are no factoring companies. The estimate
on the P ostt × T reatedi term under column (1) implies that high-fixed assets firms increased
their long-term debt ratios four percentage points more than low-fixed assets firms with the
collateral reform. This is notable when compared to the average long-term leverage of firms
with high-fixed assets usage prior to the reform, which was 1.9%. The corresponding estimate
in column (7) implies that the change in the proportion of zero-leverage firms located in areas
with no factoring companies drops by 11.4 percentage points more than the proportion of
zero-leverage firms in areas with factoring services. The effects of the reform on short-term

16
debt-taking behavior are either negligible or work in the opposite direction.

Ta b l e 1 A b o u t H e r e

Table 2 reports the results of DIDID regressions combining our two identification strate-
gies. In areas with factoring companies, high-fixed assets firms observe a 3.0 percentage points
greater increase in their long-term debt than low-fixed assets firms after the reform (see the
slope coefficient of P ost × T reatedHighF ixedAssets in column (1)). Most notably, high-fixed assets
firms located in areas with no factoring services observe a 5.2 points greater increase in their
long-term debt ratios than low-fixed assets firms located in areas with factoring (note the sum
of the slope coefficients on all three interaction terms). The effect is remarkable when compared
to the average long-term leverage of firms with high-fixed assets usage (alternatively, firms with
no access to factoring) prior to the collateral reform, which was 2.3% (or 1.4%). Estimates for
the change in the proportion of zero-leverage firms point to a similar direction (see column (3)).
They imply that the proportion of zero-leverage firms among high-fixed assets firms located in
areas with no factoring companies drops by 23.7 percentage points more than the proportion
of zero-leverage firms among low-fixed assets firms located in areas with factoring. In stark
contrast, no such effects are found for firms’ short-term debt usage (columns (2) and (4)).

Ta b l e 2 A b o u t H e r e

Our baseline estimations suggest that Ordonnance 2006-346 allowed firms operating more
fixed assets and located far from factoring companies to significantly raise their long-term
borrowing (intensive margin). The reform also lowered the fraction of zero long-term debt firms
among them (extensive margin). Neither low-fixed assets firms nor firms located near factoring
companies increased their long-term borrowing, and the proportion of zero long-term debt firms
among them remained the same.

4.2 Loan Terms and Lending Syndication

We also study whether contract terms changed following the collateral reform. If Ordonnance
2006-346 facilitated secured debt contracting, one could expect secured loans to potentially
receive more favorable terms afterwards. In performing this examination, we look at the pricing

17
and maturity terms of secured (“treated”) and unsecured (“control”) loans before and after the
reform. The DID regressions in Table 3 show that secured loans became significantly cheaper
and carried longer maturities compared to unsecured loans after the reform. In column (1), the
log all-in-drawn spread of secured loans dropped by 1.20 more than the spread of unsecured
loans, implying a 146 basis points greater reduction in mark-ups. The maturity of secured
loans increased by 94 months more than that of unsecured loans, a significant amount of time
considering the pre-reform average maturity time of 74 months (see column (2)).
In addition to debt contract terms, we also study the impact of the reform on the number of
lenders involved in secured loan transactions. Since the reform introduced rechargeable security
interests, we expect that it made it easier to set up syndicated loans, leading to an increase in
the number of lenders involved in secured lending. The estimates under column (3) in Table 3
imply that the number of lenders involved in secured loan agreements increased by one lender
with the reform relative to the number of lenders involved in unsecured loans, a notable figure
if compared to the pre-reform average of two lenders.

Ta b l e 3 A b o u t H e r e

Our contract-level evidence suggests that the 2006 collateral reform made banks more
willing to extend secured credit on more favorable terms. It also prompted secured loans to be
syndicated by a greater number of lenders. Our results imply that policies that focus on easing
collateral contracting can lead to a relaxing effect on several dimensions of corporate borrowing.

5 Demographics of New Borrowers

5.1 Attributes of Firms Accessing Credit after Ordonnance 2006-346

Firms coming into credit markets following credit reforms may not necessarily be those targeted
by policymakers, and enlarged credit access may ultimately bring undesirable economic conse-
quences (e.g., Lilienfeld-Toal et al. (2012)). This section depicts the characteristics of those firms
that report the greatest gains in credit access under Ordonnance 2006-346. We focus on firms
with no observed credit access before the reform and positive long-term leverage ratio in every
post-reform year (likely “marginal borrowers”). We name them “new borrowers.” We compare

18
Figure 5. Comparison of New Borrowers with Population in Industry New borrowers (red bars) are
those firms with zero long-term leverage ratio in every pre-reform year (2001–2005) and positive long-term
leverage in every post-reform year (2006–2009). Population in industry (blue bars) are all other companies that
operate in the same industry of the new borrowers. To calculate the figure entries, for each new borrower we
first match the variable of interest with the average value of the variable across all other firms that operate
in the same industry, using data from 2005 only. We then average separately by the new borrowers and the
industry matches. The vertical lines give the 99% confidence bands around the mean estimates. Size is total
assets in million $; Dist10Cities is the spherical distance between the firm’s location and the closest of the ten largest
French cities; Cash is cash reserves divided by total assets; Profitability is the ratio of EBIT to total assets; and
Loss Dummy is a dummy variable equal to one if Profitability is negative, else zero.

new borrowers with other firms that operate in the same industry (“population in industry”).
Figure 5 compares the new borrowers and their industry-peers across several characteristics,
based on pre-reform average statistics (compare the red and the blue bars). The figure shows
that new borrowers were smaller and located farther away from large cities. They were also
more cash-strapped. Notably, their profitability exceeded that of the average firm in the same
industry and they were also less likely to run a loss.
Figure 5 suggests that the collateral reform incentivized creditors to extend long-term fi-
nancing to small, cash-starved, profitable firms. One can argue that these firms are “desirable”
marginal borrowers that were brought into credit markets by the reform. Our evidence also
suggests that the firms that benefitted most from the reform were located far from big cities,
implying that the reform extended the reach of credit markets deeper into France’s countryside.
We examine these distributional effects in more detail below.

19
5.2 The Financing of Start-Up Firms

Prior research argues that credit market deregulation enables prospective entrepreneurs to
compete with incumbent firms for external financing (Bertrand et al. (2007), Kerr and Nanda
(2009), and Chatterji and Seamans (2012)). We study whether Ordonnance 2006-346 affected the
debt-taking behavior of French start-ups in Table 4, where we contrast high-fixed and low-fixed
assets start-up firms before and after the reform. Comparisons are based on a propensity-score
(PS) matching. To match firms, we use a logit model to estimate the probability that a firm
has a fixed assets-to-total assets ratio in the top quartile of that distribution. We use firm size,
the number of employees, and profitability as covariates. We match each of the firms belonging
to the top fixed assets-to-total assets ratio quartile with a firm belonging to other quartiles and
having the closest logit fitted probability, within the same year of incorporation.
The results in Table 4 suggest that France’s new entrepreneurs took advantage of the en-
hanced access to credit brought about by Ordonnance 2006-346. Panel A shows that, before the
reform, start-ups had an average long-term leverage of about 2.0% in their year of incorporation.
This figure increased to 3.4% after the reform. Panel B shows that, before the reform, 92% of all
start-ups had no long-term debt in their year of incorporation. This figure dropped to 71% after
the collateral reform. Consistent with our prior findings, a firm’s fixed assets usage directly con-
ditions these results. In particular, the high-fixed assets start-ups show a 26.7 percentage points
greater drop in their proportion of zero long-term debt firms than the low-fixed assets start-ups.
Indeed, start-up firms operating fewer fixed assets seem unaffected by the collateral reform.

Ta b l e 4 A b o u t H e r e

6 The Geography of Reform Effects

6.1 Effects of the Credit Reform Across Departments

Lilienfeld-Toal et al. (2012) show that credit reforms can channel credit away from firms in
poor rural areas toward firms in large cities. This potentially perverse dynamic raises serious
policy concerns. Our setting and data allow us to study the geography of distributional effects
of a reform that enlarges access to collateral.

20
Figure 6. The Long-term Leverage Distribution of French Departments Before and After the
Reform The histograms show the distribution of mean long-term leverage and the distribution of the proportion
of zero long-term leverage firms for the 96 French departments and the pre- or post-reform period. The histogram
entries are the mean of long-term leverage and the proportion of zero long-term leverage firms by department
and year, separately averaged over the pre-reform and the post-reform years.

We first study whether the 2006 collateral reform affected corporate debt-taking differentially
across French departments. Figure 6 depicts the distributions of departments’ firm mean long-
term leverage ratios (Panel A) and proportions of zero long-term leverage firms (Panel B)
before and after Ordonnance 2006-346. Before the reform, the vast majority of departments
had a mean leverage ratio between 1% and 3%, and a mean proportion of zero-leverage firms
ranging between 60% and 90%. These figures changed significantly with the 2006 reform.
Notably, the reform shifted the bulk of the mean leverage distribution to the right (see right-
side histogram in Panel A), so that after 2006 most departments registered a mean leverage ratio
between 4% and 9%. Likewise, the reform shifted the bulk of the proportion of zero-leverage firms
distribution to the left (see right-side histogram in Panel B), so that after 2006 most departments

21
registered a mean proportion of zero-leverage firms between 0% and 30%. The collateral reform
reshaped the distribution of long-term debt usage across virtually all French departments.
We next estimate the probability distribution of changes in corporate debt-taking across
different areas of the country around the reform. We do so at a granular level using spatial
point analysis. We first compute the fraction of firms in each department with a five (or more)
percentage points change in their long-term debt ratio. We then calculate the median value
of this fraction across all departments and identify the location of those departments that are
above the sample median. Using those point locations (“department centroids”), we generate
kernel estimates of the probability distribution of a change in corporate debt-taking across all
areas of the country; where “areas” are arbitrarily-small defined. We also identify the fraction
of firms in each department that never had any long-term debt before the reform, but report a
positive long-term leverage ratio every post-reform year (new borrowers). We again calculate
the median value of this fraction across all departments and identify those departments above
the median. The estimation of the likelihood that firms in an area increase their long-term
leverage above a certain threshold allows us to build “heat maps” of the change in debt-taking
around Ordonnance 2006-346.22
We show the heat maps in Figure 7. A deeper red indicates a higher probability of observing
firms that experienced large increases in debt-taking after the reform in a given locality. For
geographic context, the black-rimmed areas represent the departments containing the largest
French cities (over one million inhabitants).
We use Moran’s I statistic to examine whether the visual color clusterings in our maps are
22
Spatial point analysis allows us to map and determine whether the distribution of an event s in a geo-
graphic surface R exhibits clustering — as opposed to being random. To implement a spatial point analysis, one
must estimate the probability density function of events, p(s). The function defines the probability of observing
an event in a certain location within R and can be estimated by means of non-parametric methods; for example,
kernel estimators. In our case, the surface R represents France and is divided into a set of 60,000 contiguous
cells (grid). The events of interest are: (1) companies with large changes in mean long-term debt (5% or more);
and (2) companies that never have any long-term debt before the reform, but have positive long-term debt every
year after the reform. For each department, we calculate the fraction of these companies, and then calculate the
median value across all departments. Next, we identify the geographic coordinates of the “centroids” — points
at the center of each department, where the capital city is usually located — of those departments above the
median. Based on these coordinates, we follow Bailey and Gatrell (1995) in using kernel estimators to generate
a spatially smooth estimate of p(s) over the entire grid of cells covering R. That is, we estimate the probability
of observing the events of interest across each of the 60,000 areas of the grid covering France. Heat maps display
color-coded results of this estimation.

22
Figure 7. Long-term Debt Financing Effects of the Reform by Department The maps show those
areas in France where the fraction of firms with a five (or more) percentage points change in long-term leverage
from the pre-reform (2001–2005) to the post-reform period (2006–2009) exceeds the median across departments
(Panel A) and where the fraction of firms with no long-term debt preceding the reform, but with a positive
long-term leverage ratio after the reform exceeds the median across departments (Panel B). The more red (blue)
an area’s color in the maps, the greater (weaker) is the probability to have French departments with above
median changes in mean long-term leverage or the proportion of new borrowers. The black-rimmed departments
contain cities with more than one million inhabitants.

statistically significant.23 To do so, we construct a Z-Score based on the mean and variance of
Moran’s I as:
I − E(I)
Z-Score = p , (6)
V ar(I)
where E(I) is the expected value of Moran’s I under the null hypothesis of no spatial correlation.
In other words, it represents the case when observations of the variable of interest (e.g., the
fraction of firms with 5% (or more) changes in leverage) are randomly distributed across
departments with no identifiable pattern. V ar(I) is the variance of Moran’s I.
23
Moran’s I statistic is calculated as:
PNPN
n k=1 m=1 w(k,m) (yk − ȳ)(ym − ȳ)
I= PN PN PN (5)
( k=1 m=1 w(k,m) ) k=1 (yk − ȳ)2

where n is the total number of observations for variable y. Variable y represents either: (1) the fraction of firms
in each department with a five (or more) percentage points change in long-term leverage from the pre-reform
to the post-reform period; or (2) the fraction of firms in each department with no long-term debt preceding
the reform, but with a positive long-term leverage ratio after the reform. k and m are department indexes,
and w denotes a row standardized binary weighted matrix, in which each element w(k,m) is equal to one if
departments k and m are less than 100 miles apart from each other, else zero. We use one observation of y
for each department for a total of 96 observations. Moran’s I measures the spatial correlation between the
value of each departmental observation with those of its neighbors. If the departmental observations of y are
independent and identically distributed (iid ), then Moran’s I is asymptotically normal with expected value of
−1
n−1 .

23
Figure 8. Degree of Overlapping between the Maps in Figure 7 The graph shows the degree of
overlapping between the maps in Figure 7. The horizontal axis shows the 20 quantiles into which each probability
distribution is divided (one for each color in the heat map). The vertical axis reports the fraction of cells in
each quantile that are in common across the two maps over the total number of cells in the quantile.

Both maps in Figure 7 have positive, highly significant Moran’s Z-Scores, with the left
map yielding a Z-Score of 14.1 and the right map a Z-Score of 9.4. In words, the geographic
distribution of high values (and low values) of the credit statistics of interest is more spatially
clustered than would be expected under a random process, implying that the firms observing the
largest increases in debt-taking after the reform are physically close to each other. To corroborate
our inferences about the reform’s effect on corporate credit-taking on intensive and extensive
margins, we calculate the degree of overlapping across the two maps in Figure 7, using the kernel
estimates of the probability of occurrence of the two credit-taking events in each single cell. Each
probability distribution is divided into 20 quantiles, one for each color in the maps. For each
quantile, we calculate the fraction of cells in common across the two maps over the total number
of cells in the quantile. The higher the fraction, the larger is the degree of overlapping between the
maps. We present the results from these calculations in Figure 8. It shows that the largest degree
of overlapping occurs in the deeper red and deeper blue areas of the two maps. The correlation
between the two probability distributions is equal to 0.71 with a p-value of less than 1%.
The inference one can draw from the above spatial point geographic analysis is that Ordon-
nance 2006-346 led to significant increases in debt-taking along intensive and extensive margin
across firms located in rural areas, away from traditional credit markets; especially those in
France’s Southwest region.

24
Figure 9. Long-term Debt Gini Index Effects of the Reform by Department The maps show each
department’s long-term debt Gini index value for the pre-reform period (Panel A) and the post-reform period
(Panel B). We first calculate the Gini index in each department each year, and then average by department for pre-
reform (2001–2005) and post-reform (2006–2009) periods separately. More (less) blue departments have a high (low)
Gini index value. The black-rimmed departments are those containing cities with more than one million inhabitants.

6.2 Changes in Credit Access Inequality

Our results suggest that the collateral reform produced the largest increases in long-term debt-
taking among firms that traditionally have less access to credit markets, such as small firms and
firms in rural areas. Accordingly, they seem consistent with the idea that the reform spurred a
“democratization of credit.” We investigate this idea further by computing the Gini index of
credit access inequality for each French department. The Gini index is computed as follows:
Nk,t Nk,t
1 X X
GiniIndexk,t = 2
|LongT ermDebti,t − LongT ermDebtj,t |, (7)
2µk,t Nk,t i=1 j=1

where GiniIndexk,t is department k’s Gini index in year t, µk,t is the average long-term debt of
the firms in department k, i and j are firm indexes, and Nk,t is the number of firms in department
k. We first calculate GiniIndexk,t in each department k in each year t. We then average by
department for pre- and post-reform periods separately. The calculation uses a fixed number
of observations (the 20,000 largest firms by total assets) in each year in order to avoid outputs
stemming from changes in sample size. The Gini index ranges from zero to one. A higher value
indicates a more unequal distribution of long-term debt usage within a department.
Figure 9 shows each department’s pre- and post-reform Gini index values. Darker (lighter)
blue departments have a relatively high (low) index value. Before the reform, 91 out of the

25
96 French departments had a Gini index above 0.90, suggesting that long-term debt was very
heavily concentrated in the hands of a few large firms. Following the reform, Gini index values
dropped quite significantly (to around 0.82).24 Most notably, while all departments observed
declines in their index values, it was the rural departments that experienced the most dramatic
drops. The results are remarkable in suggesting that the collateral reform reduced inequalities in
the access to credit, with firms in the countryside experiencing the greatest gains in credit access.

7 Real Outcomes
We conclude our analysis by assessing how firms used the credit made available by the collateral
reform. If firms used their new funds to invest into profitable projects, the resulting change
in their performance could be seen as a positive consequence of the credit expansion. If, on
the other hand, the impact of the reform was simply an increase in firm indebtness, one would
worry about increased corporate default rates and other negative consequences of extending
credit to marginal firms.

7.1 Firm Growth and Performance

Table 5 shows the results of DID regressions featuring the following firm outcome variables:
fixed investment, number of employees, sales, profitability, and profit volatility. Our estimates
suggest that the firms that benefitted the most from the reform (high-fixed assets firms) invested
more and hired more workers after the reform. Indeed, those firms observed a 5.3 percentage
points higher employment growth than low-fixed assets firms (more on this below). They also
observed a 4.2 percentage points higher sales growth than the control firms. Further, high-fixed
assets firms not only raised their output more, but they also became more profitable and less
risky (have lower profit volatility) after the reform.

Ta b l e 5 A b o u t H e r e
24
It is worth discussing how one could read these Gini index figures. Recall, before 2006, nearly 90% of the
firms in the country did not have any long-term debt. A Gini index of 0.90 before 2006 is equivalent to a case
where 1% of the firms in the population held 80% of the outstanding debt, 9% of the firms held the remaining
20%, while 90% had no access to debt financing. Notably, after the reform, about 58% of firms held some debt. A
Gini index of 0.82 after 2006 would be consistent with the case where 1% of the firms held 56% of the long-term
debt in the economy, 9% held 24%, and 48% of the population held the remaining 20%.

26
7.2 Firm Survival

We also study corporate default rates. Arguably, changes in default rates following a credit re-
form may more directly measure whether it has inadvertently extended credit to risky marginal
borrowers. We use a proportional Cox hazard model to fit the number of years until a firm fails.
The model assumes that the hazard rate is given by:
 
λi,t = φt exp βP ostt × T reatedHighF
i
ixedAssets
+ γT reatedHighF ixedAssets
i + X i,t δ , (8)

where the hazard rate, λi,t , is defined as the probability of firm i failing at time t conditional on
surviving until this time. We set the failure year equal to the calendar year during which a firm’s
legal status changes from an active status to one of the failure statuses (if it does so).25 φt is the
baseline hazard rate common to all firms. The exponential function allows for cross-sectional
variations in hazard. Using a partial likelihood estimator, we obtain estimates for β, γ, and δ
without imposing any structure on φt . The estimator accounts for right-censoring of the data.
Table 6 shows that, in the absence of controls, firms operating more fixed assets experience
a 53.5 percentage points (e−0.766 ) greater decline in their failure rates than other firms after the
reform. Controlling for leverage, and thus for the fact that high-fixed assets firms increased their
debt levels more than other firms, the difference in the change in failure rates between high-
and low-fixed assets firms widens. Finally, adding profitability and profit volatility reduces the
difference again. This suggests that the high-fixed assets firms’ greater increase in profitability
and greater decline in profit volatility partially explain why their failure rates decline relative
to the failure rates of low-fixed assets firms after the passage of the collateral reform.

Ta b l e 6 A b o u t H e r e

7.3 Labor Market Effects

Table 5 suggests that firms that benefited from a reform that eased debt-taking both invested
more in fixed assets and hired more workers. This is consistent with the idea that the availability
of credit has an effect on employment through firms’ investment decisions, as labor and capital
25
The failure statuses in the AMADEUS dataset are: “default of payment,” “insolvency proceedings,” “re-
ceivership,” “bankruptcy,” “dissolved (bankruptcy),” “dissolved (liquidation),” and “in liquidation.”

27
Figure 10. Labor Market Effects of the Reform by Department The maps show those areas in France
where the proportion of firms with an increase in mean employment from the pre-reform to the post-reform
period exceeds the nation-wide median (Panel A), where the proportion of firms with an increase in mean average
wages exceeds the nation-wide third quartile (Panel B), and where the proportion of firms with an increase in
mean total factor productivity exceeds the nation-wide median (Panel C). The more red (blue) a region’s color
in the maps, the greater (weaker) is the clustering of departments with above median or third quartile changes
in the variables. The black-rimmed departments contain cities with more than one million inhabitants.

are often complimentary inputs (see Chodorow-Reich (2014)). In this setting, it is important
to ask whether a credit reform like Ordonnance 2006-346 ultimately may affect the country’s
labor market and what distributional effects may entail.
We use heat maps to characterize the geographic effects of Ordonnance 2006-346 onto
France’s labor market. While our regression analysis implies that, in general, high-fixed assets
firms hired more workers after the reform, we now trace the geographic distribution of the impact
of the credit reform on labor demand. We also map the impact of the reform on wages and firm
productivity. We use total factor productivity to measure the efficient use of production inputs
(fixed capital and labor). Figure 10 shows that the largest increases in employment (Panel
A), wages (Panel B), and productivity (Panel C) mostly took place in those areas in France
which also observed the largest reform-induced increases in debt-taking (as shown in Figure 7).
Moran’s I statistic suggests statistically significant clustering in those areas experiencing the
most pronounced labor effects within each of the three maps. Our analysis suggests that by
opening up access to credit, the 2006 collateral reform appears to have spurred improvements
(e.g., higher productivity) in France’s labor markets, particularly in those regions far away from
the large metropolitan areas.

28
7.4 Capital Allocation Efficiency

Our evidence thus far suggests that the reform made creditors more willing to lend to profitable,
low-risk firms. The ultimate question is whether this change led to a more efficient allocation
of capital in the economy. We use two approaches to tackle this question.
First, we follow Wurgler’s (2000) argument that an efficient capital allocation process implies
that corporate investment increases in sectors with better growth opportunities and decreases
in sectors with poorer opportunities. Using industry value added as proxy for growth oppor-
tunities, we estimate the sensitivity of investment growth to value added growth as a way to
gauge capital allocation efficiency. We study how this sensitivity changes from the pre- to the
post-reform period. We do this running the following regression on industry-level data:

InvestmentGrowths,t = α + βV alueAddedGrowths,t + εs,t , (9)

where InvestmentGrowths,t is the percentage change in industry s’s gross fixed capital forma-
tion from year t − 1 to year t, and V alueAddedGrowths,t is the percentage change in value
added over the same period. Using pre-reform data, Table 7 reports that the elasticity between
investment growth and value added growth (β) was 0.48. Using post-reform data, the elasticity
jumped to 0.85. As benchmark, we note that the average elasticity for other Western European
Civil Laws countries is 0.72 (Wurgler (2000)). Based on Wurgler’s (2000) interpretation of these
elasticities, the reform led to an economically significant increase in capital allocation efficiency.

Ta b l e 7 A b o u t H e r e

Second, we use the external financial dependence measure of Rajan and Zingales (1998)
as alternative proxy for capital allocation efficiency. This measure captures an industry’s
technology-driven demand for external financing. In an efficient economic system, capital should
be directed toward those sectors that need more external funding. If the reform relaxed credit
constraints, we should observe significantly larger increases in borrowing among firms operating
in sectors highly dependent on external financing. To test this hypothesis, we split our sample
into two groups, one including sectors with a high dependence on external financing (proxy
value above the median) and one including sectors with a low dependence (below the median).
Next, we split the firms in the subsamples into “treated” and “control” firms, where treatment

29
status is assigned to firms with a fixed assets-to-total assets ratio within the top quartile
in 2005. Control status is assigned to propensity-score (PS) matched firms.26 The resulting
categories are covariate-balanced and of equal sizes.
Table 8 reports that firms with a high dependence on external financing observed greater
reform-induced increases in their long-term leverage than all other firms. While the mean long-
term leverage ratio of highly-dependent firms increased by 6.3 percentage points (see bottom
of column (1)), the long-term leverage ratio of firms with a weaker dependence increased by
only 3.6 points (column (4)). The difference is statistically and economically significant. More
importantly, the difference is larger for the reform-treated firms. Highly-dependent–high-fixed
assets firms observed a mean increase in their long-term leverage of 7.5 percentage points.

Ta b l e 8 A b o u t H e r e

Taken together, our tests suggest that Ordonnance 2006-346 helped the French economy
move towards allocating more capital to high-value added sectors as well as sectors that require
more external financing. These findings point to positive welfare gains produced by the 2006
collateral reform in terms of reducing allocative credit distortions.

8 Robustness

8.1 Self-Selection, Parallel Trends, and Outcome Autocorrelation

The validity of DID tests rests on a number of assumptions, one of which is that firms cannot
self-select into treatment. In our context, this implies that we need to rule out that firms manage
their assets in such a way that they seek to benefit from changing security laws. Untabulated
results show that it is unlikely that firms do so. Over the 2001–2005 period, a mere 6.8% of
firms moved from any of the lower three fixed assets quartiles into the highest fixed assets
quartile. Notably, the 2001–2005 migration rates do not differ from migration rates calculated
26
To create the PS matched sample, we use a logit model to estimate the probability that a firm has a fixed
assets-to-total assets ratio in the top quartile of that asset distribution. We estimate the model within each
industry, using data from 2005. As covariates, we use firm size, profitability, and leverage. We match each firm
belonging to the top fixed assets-to-total assets ratio quartile with that within-industry counterpart belonging
to the other quartiles and having the closest logit-model fitted probability.

30
over other five year-periods within our sample period.
Treated and control firms should also display parallel trends in outcome variables of interest
before the reform. We have shown that this is the case in Figures 2 and 3, where neither high-
or low-fixed assets firms observe any significant trends in total or long-term leverage during
the period preceding the reform. Untabulated statistical tests confirm the above inferences.
Finally, Bertrand et al. (2004) show that autocorrelation in the outcome variable can create
upward-biased inference levels in DID tests. As a robustness check, we separately average each
variable analyzed over the pre- and post-reform periods and repeat our DID regressions us-
ing the collapsed data. We report these tests in Table 9. Results under columns (2) and (5) of
Table 9 are virtually identical to those in columns (1) and (4), respectively, where we reproduce
the baseline results of Table 1.

Ta b l e 9 A b o u t H e r e

8.2 Placebo Tests

8.2.1 Public Firms

French public firms only take on secured credit facilities under exceptional circumstances (e.g.,
when they are financially distressed). We use this insight to replicate our tests on public firms
data, expecting that their borrowing is less affected by the reform than the borrowing of private
firms. While public firms experienced a slight increase in long-term leverage from the pre- to
the post-reform period, there is no evidence to suggest that the increase is driven by high-fixed
assets firms. In fact, if anything, results work in the opposite direction. While firms in the
low-fixed assets group raised their long-term leverage, firms in the high-fixed assets group
significantly cut their long-term leverage (see columns (3) and (6) of Table 9).

8.2.2 Other Civil Law Countries

One concern with our results is that changes affecting firms across Europe in 2006 (and not
necessarily Ordonnance 2006-346 ) may explain the growth in debt taking by high-fixed assets
firms in France. As a final falsification test, we repeat our analysis using data from European
Civil Law countries that already had reformed their security laws at the start of our sample pe-

31
Figure 11. Evolution of Mean Long-term Leverage and the Proportion of Zero Long-term Lever-
age Firms in the Placebo Countries The figure shows the evolution of mean long-term leverage (upper
panels) and the proportion of zero long-term leverage firms (lower panels) for each placebo country and fixed
assets quartile (Q) over the 2001–2009 period. The placebo countries are Belgium, Italy, Spain, and Portugal.
Q1, Q2, Q3, and Q4 indicate the first, second, third, and fourth fixed assets quartile, respectively.

riod: Belgium (a neighbor with close economic ties), Italy, Spain, and Portugal. Figure 11 shows
that none of the placebo countries displays effects similar to those found in France. Untabulated
regressions using data from the placebo countries confirm these inferences.

9 Concluding Remarks
We examine France’s Ordonnance 2006-346 to assess how reforms that ease credit access by en-
hancing collateral menus change the demographics of corporate debt utilization. We show that
a large amount of the newly-created debt went to small, profitable firms located in rural areas.
By easing access to credit, the reform produced notable reductions in the Gini index of credit
access inequality across the country. Our results also suggest that firms used their new debt
in welfare-enhancing ways. The firms that benefitted the most from the collateral reform grew
more profitable and were less likely to fail. The reform also appears to have affected France’s
labor market, allowing firms to hire more workers, increase their average wage, and improve
their productivity. Our evidence suggests that Ordonnance 2006-346 led to a “democratization”
of corporate credit access.
Our study shows that the derogation of the Napoleonic security code — and the embodied
notion of possessory ownership — led to far reaching changes in the make up of investment and

32
growth in the French economy. It prompted an increase in the elasticity between investment
and value added. It also allowed for more borrowing by firms operating in industries heavily
reliant on external financing. All combined, our results point to a significant improvement in
capital allocation efficiency in the economy. This is an important effect from a policymaking
perspective since reforms of this nature may require fewer resources to implement and maintain
than reforms that focus, for example, on law enforcement and judicial intervention.

33
References
Ancel, M.-E., 2008. Recent reform in France: The renaissance of a civilian collateral regime. In:
Dahan, F., Simpson, J. (Eds.), “Secured transaction reform and access to credit.” Edward
Elgar Publishing, 259-272.

Assunção, J., Benmelech, E., Silva, F., 2014. “Repossession and the democratization of credit.”
Review of Financial Studies 27, 2661-2689.

Bailey, T. C., Gatrell, A. C., 1995. “Interactive Spatial Data Analysis.” Routledge.

Benmelech, E., Moskowitz, T. J., 2010. “The political economy of financial regulation: Evidence
from U.S. state usury laws in the 19th century.” Journal of Finance 65, 1029-1073.

Bertrand, M., Duflo, E., Mullainathan, S., 2004. “How much should we trust differences-in-
differences estimates?” Quarterly Journal of Economics 119, 249-275.

Bertrand, M., Schoar, A., Thesmar, D., 2007. “Banking deregulation and industry structure:
Evidence from the French banking reforms of 1985.” Journal of Finance 62, 597-628.

Boughida, S., Levesque, G., Roux, J., 2011. “Lending and taking security in France: An
overview.” Available from http://uk.practicallaw.com/7-501-4362.

Calomiris, C. W., Larrain, M., Liberti, J., Sturgess, J., 2017. “How collateral laws shape lending
and sectoral activity.” Journal of Financial Economics 123, 163-188.

Campello, M., Larrain, M., 2016. “Enlarging the contracting space: Collateral menus, access
to credit, and economic activity.” Review of Financial Studies 29, 349-383.

Cerqueiro, G., Ongena, S., Roszbach, K., 2016. “Collateralization, bank loan rates and moni-
toring.” Journal of Finance 71, 1295-1322.

Chatterji, A. K., Seamans, R. C., 2012. “Entrepreneurial finance, credit cards, and race.”
Journal of Financial Economics 106, 182-195.

Chodorow-Reich, G., 2014. “The employment effects of credit market disruptions: Firm-level
evidence from the 2008-9 financial crisis.” Quarterly Journal of Economics 129, 1-59.

Coval, J. D., Moskowitz, T. J., 2001. “Home bias at home: Local equity preference in domestic
portfolios.” Journal of Finance 54, 2045-2073.

Davydenko, S. A., Franks, J., 2008. “Do bankruptcy codes matter? A study of defaults in
France, Germany, and the U.K.” Journal of Finance 63, 565-608.

Demirgüç-Kunt, A., Levine, R., 2008 “Finance and economic opportunity.” World Bank Policy
Research Paper 4468.

Djankov, S., McLiesh, C., Shleifer, A., 2007. “Private credit in 129 countries.” Journal of
Financial Economics 84, 299-329.

34
Garmaise, M., Moskowitz, T. J., 2004. “Confronting information asymmetries: Evidence from
real estate markets.” Review of Financial Studies 17, 405-437.

Haimo, S. H., 1983. “A practical guide to secured transactions in France.” Tulane Law Review
58, 1163-1205.

Herbet, J., Sabbah, C., 2006. “Will secured lending in France benefit from the recent overhaul
of civil code provisions relating to security interests?” International Business Law Journal
6, 853-859.

Ivashina, V., Scharfstein, D., 2010. “Loan syndication and credit cycles.” American Economic
Review: Papers & Proceeds 100, 57-61.

Kerr, W. R., Nanda, R., 2009. “Democratizing entry: Banking deregulation, financing con-
straints, and entrepreneurship.” Journal of Financial Economics 94, 124-149.

Keys, B. J., Mukherjee, T., Seru, A., Vig, V., 2010. “Did securitization lead to lax screening?
Evidence from subprime loans.” Quarterly Journal of Economics 125, 307-362.

La Porta, R., Lopez-de-Silanes, F., Shleifer, A., Vishny, R., 1998. “Law and finance.” Journal
of Political Economy 106, 1113-1155.

Leaseurope., 2009. “Annual Survey 2009.”

Lilienfeld-Toal, U., Mookherjee, D., Visaria, S., 2012. “The distributive impact of reforms in
credit enforcement: Evidence from Indian debt recovery tribunals.” Econometrica 80,
497-558.

Lofstedt, C., Baum, A., 1993. “A comparative study of commercial leasing structures in selected
European countries, and in the USA.” Centre for European Property Research.

Mian, A., Sufi, A., 2009. “The consequences of mortgage credit expansion: Evidence from the
U.S. mortgage default crisis.” Quarterly Journal of Economics 124, 1449-1496.

Mian, S. L., Smith, C. W., 1992. “Accounts receivable management policy: Theory and evi-
dence.” Journal of Finance 47, 169-200.

Omar, P., 2007. “Updating the framework for asset security in France: The reforms of 2006.”
Journal of Comparative Law 2, 189-209.

Plantin, G., Thesmar, D., Tirole, J., 2013. “Reforming French bankruptcy law.” Notes du
Conseil D’Analyse Économique 7, 1-12.

Rajan, R., 2009. “Rent preservation and the persistence of underdevelopment.” American
Economic Journal: Macroeconomics 1, 178-218.

Rajan, R., Zingales, L., 1998. “Financial dependence and growth.” American Economic Review
88, 559-586.

35
Rajan, R., Zingales, L., 2003. “Saving capitalism from the capitalists: Unleashing the power of
financial markets to create wealth and spread opportunity.” Princeton University Press.

Renaudin, M., 2013. “The modernisation of French secured credit law: Law as a competitive
tool in global markets.” International Company and Commercial Law Review 24, 385-392.

Schmalz, M. C., Sraer, D. A., Thesmar, D., 2017. “Housing collateral and entrepreneurship.”
Journal of Finance 72, 99-132.

Smith, J. K., Schnucker, C., 1994. “An empirical examination of organizational structure: The
economics of the factoring decision.” Journal of Corporate Finance 1, 119-138.

Vernimmen, P., Quiry, P., Le Fur, Y., 2014. “Finance D’Enterprise.” Dalloz-Sirey.

Vig, V., 2013. “Access to collateral and corporate debt structure: Evidence from a natural
experiment.” Journal of Finance 68, 881-928.

White Clark Group, 2015. “Global Leasing Report.” White Clarke.

White, M., 2007. “Bankruptcy reform and credit cards.” Working Paper, National Bureau of
Economic Research.

Wurgler, J., 2000. “Financial markets and the allocation of capital.” Journal of Financial
Economics 58, 187-214.

36
 Table 1
Effect of the Reform on Leverage: DID Regressions Using Fixed Assets or Factoring Companies as Identification Variable
The table shows the results from the following regression:

Yi,t = αi + αk,t + βP ostt × T reatedIDStrategy

i + Xi,t γ + εi,t ,

where Yi,t is long-term leverage, short-term leverage, or dummy variables indicating whether the former quantities are zero. Long-term leverage is
long-term debt scaled by total assets; short-term leverage is short-term debt scaled by total assets. Dummy No Long-term (Short-term) Debt is
a dummy variable equal to one if long-term (short-term) debt is zero. Post is a dummy variable equal to one for years greater or equal to 2006.
Treated IDStrategy is equal to Treated HighF ixedAssets in the columns labelled “High Fixed Assets” and to Treated N oF actoringAvailable in the columns
labelled “No Factoring Available.” Treated HighF ixedAssets is a dummy variable equal to one for firms whose fixed assets-to-total assets ratio is in the
top quartile. Treated N oF actoringAvailable is a dummy variable equal to one for firms registered in postal code areas in which there are no factoring
companies. Xi,t is a vector of control variables, including size, age, employees, and profitability. Size is the log of total assets; Age is the log of
the current year minus the year of incorporation; Employees is the log of the number of employees; and Profitability is the ratio of earnings before
interest and taxes to total assets. β and γ are free parameters. αi and αk,t indicate firm- and interacted year-industry fixed effects. εi,t is the residual.
T-statistics (in parentheses) are calculated from standard errors clustered at the firm-level.

ID Strategy: High Fixed Assets ID Strategy: No Factoring Available

Intensive Margin Extensive Margin Intensive Margin Extensive Margin

37
Dummy No Dummy No Dummy No Dummy No
Long-term Short-term Long-term Short-term Long-term Short-term Long-term Short-term
Leverage Leverage Debt Debt Leverage Leverage Debt Debt
Post × Treated 0.040*** –0.002 –0.123*** 0.010** 0.010*** 0.003** –0.114*** 0.004
(27.80) (–1.63) (–17.08) (2.42) (10.53) (2.43) (–18.00) (1.11)
Size 0.011*** 0.012*** –0.063*** –0.041*** 0.010*** 0.012*** –0.059*** –0.042***
(12.11) (9.26) (–13.55) (–10.42) (10.76) (9.24) (–12.89) (–10.54)
Age –0.004*** 0.003* 0.087*** –0.038*** –0.044*** –0.136*** 0.097*** 0.120***
(–3.20) (1.93) (10.96) (–5.93) (–13.68) (–30.78) (5.96) (8.31)
Employees –0.002* 0.002 –0.014*** –0.027*** –0.004** 0.004** 0.081*** –0.037***
(–1.67) (1.39) (–3.26) (–7.26) (–2.47) (1.97) (10.26) (–5.75)
Profitability –0.045*** –0.136*** 0.104*** 0.120*** 0.000 0.001 –0.017*** –0.026***
(–14.04) (–30.84) (6.41) (8.40) (–0.45) (1.27) (–3.91) (–7.01)
R-squared 0.08 0.01 0.24 0.01 0.05 0.01 0.25 0.01
Observations 173,522 173,522 173,522 173,522 172,200 172,200 172,200 172,200
Firm FE Yes Yes Yes Yes Yes Yes Yes Yes
Year-Industry FE Yes Yes Yes Yes Yes Yes Yes Yes
∗∗∗ ∗∗ ∗
, , and indicate statistical significance at the 99%, 95%, and 90% confidence levels, respectively.
Table 2
Effect of the Reform on Leverage: DIDID Regressions Using Fixed Assets and Lack of Factoring
Companies as Identification Variables
The table shows the results from the following regression:

Yi,t = αi + αk,t + βP ostt × T reatedHighF

i
ixedAssets
+ δP ostt × T reatedN
i
oF actoringAvailable

+γP ostt × T reatedHighF

i
ixedAssets
× T reatedN
i
oF actoringAvailable
+ Xi,t η + εi,t ,

where Yi,t is long-term leverage, short-term leverage, or dummy variables indicating whether the former quantities
are zero. Long-term leverage is long-term debt scaled by total assets; short-term leverage is short-term debt
scaled by total assets. Dummy No Long-term (Short-term) Debt is a dummy variable equal to one if long-
term (short-term) debt is zero. Post is a dummy variable equal to one for years greater or equal to 2006.
Treated HighF ixedAssets is a dummy variable equal to one for firms whose fixed assets-to-total assets ratio is in
the top quartile. Treated N oF actoringAvailable is a dummy variable equal to one for firms registered in postal
code areas in which there are no factoring companies. Xi,t is a vector of control variables, including size, age,
employees, and profitability. Size is the log of total assets; Age is the log of the current year minus the year of
incorporation; Employees is the log of the number of employees; and Profitability is the ratio of earnings before
interest and taxes to total assets. β, δ, γ, and η are free parameters. αi and αk,t indicate firm- and interacted
year-industry fixed effects. T-statistics (in parentheses) are calculated from standard errors clustered at the
firm-level.

Dependent Variable
Intensive Margin Extensive Margin
Dummy No Dummy No
Long-term Short-term Long-term Short-term
Leverage Leverage Debt Debt
Post × Treated HighF ixedAssets 0.030*** –0.004* –0.099*** 0.019***
(12.91) (–1.85) (–8.16) (2.59)
Post × Treated N oF actoringAvailable 0.006*** 0.002 –0.105*** 0.007
(6.44) (1.62) (–14.44) (1.65)
Post × Treated HighF ixedAssets × 0.016*** 0.003 –0.033** –0.014
Treated N oF actoringAvailable (5.64) (1.13) (–2.27) (–1.57)
Size 0.011*** 0.012*** –0.062*** –0.042***
(11.92) (9.19) (–13.56) (–10.48)
Age –0.044*** –0.136*** 0.096*** 0.120***
(–13.72) (–30.77) (5.95) (8.31)
Employees –0.004*** 0.004** 0.084*** –0.037***
(–3.00) (1.99) (10.57) (–5.78)
Profitability –0.001 0.002 –0.014*** –0.026***
(–1.56) (1.32) (–3.18) (–7.08)
R-squared 0.08 0.01 0.25 0.01
Observations 172,200 172,200 172,200 172,200
Firm FE Yes Yes Yes Yes
Year-Industry FE Yes Yes Yes Yes
∗∗∗ ∗∗ ∗
, , and indicate statistical significance at the 99%, 95%, and 90% confidence levels, respectively.

38
Table 3
Effect of the Reform on Loan Contracting Terms: DID Regressions
The table shows the results from the following regression:

Zi,l,t = αk,t + βP ostt × T reatedSecured

i,l,t + Xi,t δ + Wi,l,t γ + εi,l,t ,

where Zi,l,t is either the loan spread, defined as the log of the sum of a loan’s coupon and annual fees scaled by its
nominal value minus the six month LIBOR rate (in basis points; LoanSpread ), the loan time-to-maturity, defined
as the log of the difference between the loan’s maturity date and its initiation date (in months; LoanMaturity),
or the log of the number of lenders involved in a loan (NumberLenders). Post is a dummy variable equal to one
for years greater or equal to 2006. Treated Secured is a dummy variable equal to one for secured loans. Xi,t is a
vector of firm-specific controls. Wi,l,t is a vector of loan contract-specific controls. Size is the log of total assets;
Age is the log of the current year minus the year of incorporation; and Profitability is the ratio of earnings
before interest and taxes to total assets. Rating is a dummy variable equal to one if the firm taking out the
loan is rated, else zero. LoanAmount is the log of the notional value of the loan; and LoanType is a dummy
variable equal to one for term loans, else zero. We also add the other endogenous variables to the controls. β, δ,
and γ are free parameters. αk,t indicates interacted year-industry fixed effects. εi,l,t is the residual. T-statistics
(in parentheses) are calculated from standard errors clustered at the firm-level.

Dependent Variable
Loan Spread Loan Maturity Number Lenders
Post × TreatedSecured –1.200*** 0.782*** 0.410***
(–3.33) (7.34) (2.60)
Size 0.048 –0.028*** 0.090*
(1.40) (–3.09) (1.81)
Age –0.031* –0.006 0.079
(–1.76) (–0.77) (1.40)
Profitability 0.008 0.183*** 0.422
(0.09) (5.17) (0.86)
Rating –0.021 0.178 –0.247**
(–0.07) (1.44) (–2.30)
LoanAmount –0.093*** –0.001 0.138***
(–6.43) (–0.04) (3.92)
LoanType 0.309*** 0.038 –0.204***
(7.86) (1.11) (–5.64)
LoanMaturity 0.318*** 0.104
(3.93) (1.22)
NumberLenders 0.084 0.042
(1.34) (1.13)
LoanSpread 0.225*** 0.147
(4.88) (1.44)
TreatedSecured 1.627*** –0.503*** –0.662***
(17.36) (–3.65) (–7.39)
R-squared 0.58 0.51 0.64
Observations 407 407 407
Year-Industry FE Yes Yes Yes
∗∗∗ ∗∗ ∗
, , and indicate statistical significance at the 99%, 95%, and 90% confidence levels, respectively.

39
Table 4
Effect of the Reform on the Leverage of Start-Ups: PS-Matched ANOVA Tests
The table gives descriptive statistics on the long-term leverage of start-ups in their year of incorporation. The
table considers the whole sample (“All”) and the treated- and control-firm subsamples. We study two attributes
of long-term leverage: the average (Panel A) and the proportion of zero long-term leverage firms (Panel B).
Long-term leverage (“LongTermLeverage”) is defined as long-term debt scaled by total assets. We create the
sample of treated (“High Fixed Assets Group”) and matched control firms (“Low Fixed Assets Group”) using
Propensity-score (PS) matching. To do so, we use a logit model to estimate the probability that a firm has a
fixed assets-to-total assets ratio in the top quartile of that asset distribution. As covariates, we use firm size,
the number of employees, and profitability. We match each firm belonging to the top fixed assets-to-total assets
ratio quartile with the one firm belonging to the other quartiles that has the closest fitted probability. The table
reports averages over the annual pre-reform period (2001–2005) averages, averages over the annual post-reform
period (2006–2009) averages, and their differences. The final column shows the differences in attributes across
the high and low fixed assets group. Standard errors are clustered at the firm level (whenever possible).

Fixed Assets Quartile Diff.

Year All High Low High–Low
Panel A: Mean Long-term
Leverage Ratio
Mean 2001–2005 (1) 0.020 0.024 0.017
Mean 2006–2009 (2) 0.034 0.054 0.013
Diff. (2)–(1) 0.014** 0.030*** –0.004 0.034**

Panel B: Proportion No
Long-term Leverage Firms (%)
Mean 2001–2005 (1) 0.916 0.909 0.923
Mean 2006–2009 (2) 0.712 0.572 0.853
Diff. (2)–(1) –0.204*** –0.337*** –0.070 –0.267***
∗∗∗ ∗∗ ∗
, , and indicate statistical significance at the 99%, 95%, and 90% confidence levels, respectively.

40
Table 5
Effect of the Reform on Firm Performance: DID Regressions
The table shows the results from the following regression:

Yi,t = αi + αk,t + βP ostt × T reatedHighF

i
ixedAssets
+ Xi,t γ + εi,t ,

where Yi,t represents firm investment, employees, sales, profitability, or profit volatility. Investment is the sum of
the change in tangible fixed assets from the prior fiscal year end to the current fiscal year end and depreciation,
scaled by the average of total assets over the two fiscal year ends. Employees is the log of the number of
employees. Sales is the log of sales. Profitability is earnings before interest and taxes scaled by total assets.
ProfitVolatility is the standard deviation of Profitability over the most recent four fiscal years, including the
most recent one. We set ProfitVolatility equal to missing if it is based on fewer than three observations. Post is
a dummy variable equal to one for years greater or equal to 2006. Treated HighF ixedAssets is a dummy variable
equal to one for firms whose fixed assets-to-total assets ratio is in the top quartile. Xi,t is a vector of control
variables, including size, age, and total leverage. Size is the log of total assets; Age is the log of the current
year minus the year of incorporation; and TotalLeverage is total debt scaled by total assets. β and γ are free
parameters. αi and αk,t indicate firm- and interacted year-industry fixed effects. εi,t is the residual. T-statistics
(in parentheses) are calculated from standard errors clustered at the firm-level.

Dependent Variable
Profit
Investment Employees Sales Profitability Volatility
Post × Treated HighF ixedAssets 0.004*** 0.053*** 0.042*** 0.006*** –0.002***
(5.08) (7.11) (5.71) (5.45) (–4.04)
Size 0.045*** 0.502*** 0.778*** 0.011*** –0.010***
(48.28) (58.55) (101.22) (11.59) (–16.82)
Age –0.044*** 0.168*** 0.211*** 0.021*** –0.018***
(–21.59) (18.20) (24.48) (13.91) (–10.57)
TotalLeverage –0.020*** –0.009 –0.220*** –0.106*** 0.012***
(–7.36) (–0.58) (–13.41) (–36.24) (8.28)
R-squared 0.02 0.46 0.59 0.01 0.04
Observations 209,634 174,806 230,615 228,780 158,223
Firm FE Yes Yes Yes Yes Yes
Year-Industry FE Yes Yes Yes Yes Yes
∗∗∗ ∗∗ ∗
, , and indicate statistical significance at the 99%, 95%, and 90% confidence levels, respectively.

41
Table 6
Effect of the Reform on Failure Rates: DID Regressions
The table shows the results from the following proportional Cox hazard model:
 
λi,t = φt exp βP ostt × T reatedHighF
i
ixedAssets
+ γT reatedHighF
i
ixedAssets
+ Xi,t δ ,

where the hazard rate, λi,t , is the probability of firm i failing at time t conditional on surviving until then,
and φt is the “baseline” hazard rate common to all firms. We set a firm’s failure year to the calendar year
in which its legal status changes from an active status to one of the failure statuses: “default of payment,”
“insolvency proceedings,” “receivership,” “bankruptcy,” “dissolved (bankruptcy),” “dissolved (liquidation),” and
“in liquidation.” We exclude a firm from the analysis if it is unclear whether it became inactive for performance
reasons, that is, when legal status changes to: “inactive (no precision),” “unknown,” and “dissolved.” Post is
a dummy variable equal to one for years greater or equal to 2006. Treated HighF ixedAssets is a dummy variable
equal to one for firms whose fixed assets-to-total assets ratio is in the top quartile. Xi,t is a vector of control
variables, including total leverage, profitability, and profit volatility. TotalLeverage is the sum of short-term
and long-term debt to total assets. Profitability is the ratio of earnings before interest and taxes to total
assets. ProfitVolatility is the standard deviation of Profitability over the most recent four fiscal years. We set
ProfitVolatility equal to missing if it is based on fewer than three observations. β, γ, and δ are free parameters.
T-statistics (in parentheses) are calculated from standard errors clustered at the firm-level.

Dependent Variable = Time-to-Failure

Post × Treated HighF ixedAssets –0.766*** –0.848*** –0.866***
(–4.25) (–4.69) (–4.21)
TotalLeverage 1.184*** 0.466***
(7.62) (2.68)
Profitability –4.959***
(–13.30)
ProfitVolatility 2.358***
(2.82)
Treated HighF ixedAssets 0.442*** 0.411** 0.686***
(2.62) (2.42) (3.58)
R-squared 0.01 0.01 0.03
Observations 182,249 182,222 129,946
∗∗∗ ∗∗ ∗
, , and indicate statistical significance at the 99%, 95%, and 90% confidence levels, respectively.

42
Table 7
Effect of the Reform on Capital Allocation
The table shows the results from the following panel-data regression:

InvestmentGrowths,t = α + βV alueAddedGrowths,t + εs,t ,

where InvestmentGrowths,t is the percentage change in gross fixed capital formation for industry s from
year t − 1 to year t, and V aluedAddedGrowths,t is the percentage change in value added over the same time
period. We perform this regression using pre-reform (2001–2005) and post-reform (2006–2007) data. α and β
are free parameters, and εs,t is the residual. In addition to the parameter estimates and their inference levels
(T-Statistic), the table shows the number of observations (Obs) and the R-squared (R-squared) per regression,
and it also reports the difference in the slope coessfficient across the two periods.

Year Obs Elasticity (β) T-Statistic R-Squared

2001–2005 (1) 517 0.481 5.55 0.06
2006–2007 (2) 217 0.845 6.46 0.16
Diff. (2)–(1) 0.364 2.20

43
 Table 8
Effect of the Reform on Firms with a High and Low External Financial Dependence: PS-Matched ANOVA Tests
The table shows the mean long-term leverage ratio of all firms (“All”), high fixed assets-, and low fixed-assets firms, separately for those with a high
or low dependence on external financing. Long-term leverage (“LongTermLeverage”) is defined as long-term debt scaled by total assets. Fixed assets
(“FixedAssets”) is defined as fixed assets scaled by total assets. We proxy for external financial dependence (“ExternalFinancialDependence”) by
calculating the median proportion of capital expenditure that are not financed by cash flows from operations for U.S. public firms. We perform these
calculations using COMPUSTAT data over the period from 1975–2005. We classify as firms with a high (low) external financial dependence those
operating in industries with an above (below) median ExternalFinancialDependence value. We create the sample of treated and matched control firms
using Propensity-score (PS) matching. To do so, we use a logit model to estimate the probability that a firm has a fixed assets-to-total assets ratio in
the top quartile of that asset distribution. We estimate the model within each industry using only data from the year directly preceding the reform
(2005). As covariates, we use firm size, profitability, and leverage. We match each firm belonging to the top fixed assets-to-total assets ratio quartile
with the one firm belonging to the other quartiles that has the closest fitted probability. At the bottom, we report averages over the pre-reform period
means (2001–2005) and the post-reform period means (2006–2009) and their differences. The penultimate (final) column shows the differences in
means across all firms (treated firms) with a high and all with a low external financial dependence. Standard errors are clustered at the firm level.

External Financial Dependence High–Low External Financial

High Low Dependence Across

44
Year All (1) Treated (2) Controls (3) All (4) Treated (5) Controls (6) All (1)–(4) Treated (2)–(5)
2001 0.017 0.022 0.013 0.018 0.022 0.016 –0.001 0.001
2002 0.016 0.020 0.012 0.018 0.023 0.016 –0.003 –0.003
2003 0.015 0.018 0.011 0.018 0.023 0.015 –0.003** –0.005*
2004 0.015 0.017 0.013 0.016 0.022 0.013 –0.001 –0.005**
2005 0.018 0.020 0.017 0.017 0.023 0.015 0.001 –0.003
2006 0.067 0.079 0.055 0.048 0.062 0.042 0.018*** 0.017***
2007 0.074 0.088 0.061 0.051 0.065 0.043 0.024*** 0.022***
2008 0.087 0.103 0.071 0.059 0.073 0.052 0.028*** 0.030***
2009 0.097 0.114 0.080 0.065 0.082 0.056 0.033*** 0.032***
Mean 2001–2005 (1) 0.019 0.022 0.015 0.018 0.024 0.015 0.001*** –0.002***
Mean 2006–2009 (2) 0.082 0.097 0.065 0.054 0.070 0.046 0.028* 0.027**
Diff. (2)–(1) 0.063*** 0.075*** 0.049*** 0.036*** 0.046*** 0.031*** 0.027*** 0.029***
∗∗∗ ∗∗ ∗
, , and indicate statistical significance at the 99%, 95%, and 90% confidence levels, respectively.
 Table 9
Robustness Tests: DID Regressions Using Fixed Assets as Identification Variable
The table shows the results from the following regression:
ixedAssets
Yi,t = αi + αk,t + βP ostt × T reatedHighF
i + Xi,t γ + εi,t ,

where Yi,t is long-term leverage in the columns titled “Intensive Margin” and a dummy variable equal to one if long-term leverage is zero in the
columns titled “Extensive Margin.” Post is a dummy variable equal to one for years greater or equal to 2006. Treated HighF ixedAssets is a dummy
variable equal to one for firms whose fixed assets-to-total assets ratio is in the top quartile. Xi,t is a vector of control variables, including size, age,
employees, and profitability. Size is the log of total assets; Age is the log of the current year minus the year of incorporation; Employees is the log of
the number of employees; and Profitability is the ratio of earnings before interest and taxes to total assets. β and γ are free parameters. αi and αk,t
indicate firm- and interacted year-industry fixed effects. εi,t is the residual. T-statistics (in parentheses) are calculated from standard errors clustered
at the firm-level. In the columns titled “Base Case (Reference),” we repeat the base-case estimation results from Table 1. In the columns titled “Pre-
and Post- Collapsed Data (Robustness Test),” we separately average each firm’s analysis variable values over the pre- and the post-reform period
before estimation. In the columns titled “Public Firms (Placebo Tests),” we repeat the base-case estimation on the sample of French public firms.

Intensive Margin Extensive Margin

Pre- and Post- Pre- and Post-
Collapsed Collapsed
Base Case Data Public Firms Base Case Data Public Firms

45
(Reference) (Robustness Test) (Placebo Test) (Reference) (Robustness Test) (Placebo Test)
P ost×T reatedHighF ixedAssets 0.040*** 0.042*** –0.042** –0.123*** –0.124*** 0.019
(27.80) (28.43) (–2.02) (–17.08) (–17.60) (0.45)
Size 0.011*** 0.013*** –0.017 –0.063*** –0.073*** –0.057*
(12.11) (9.45) (–0.58) (–13.55) (–9.77) (–1.83)
Age –0.004*** –0.003 0.028 0.087*** 0.130*** –0.129
(–3.20) (–1.66) (0.55) (10.96) (10.85) (–1.31)
Employees –0.002* 0.000 0.024 –0.014*** –0.021*** –0.024
(–1.67) (0.12) (1.14) (–3.26) (–2.90) (–0.78)
P rof itability –0.045*** –0.068*** –0.093 0.104*** 0.249*** 0.075
(–14.04) (–9.17) (–1.04) (6.41) (6.15) (0.75)
R-squared 0.08 0.09 0.01 0.24 0.18 0.13
Observations 173,522 48,962 1,950 173,522 48,962 1,950
Firm FE Yes Yes Yes Yes Yes Yes
Year-Industry FE Yes Yes Yes Yes Yes Yes
∗∗∗ ∗∗ ∗
, , and indicate statistical significance at the 99%, 95%, and 90% confidence levels, respectively.