Journal of Empirical Finance: Yulia Davydova, Vladimir Sokolov

Journal of Empirical Finance 29 (2014) 247265
Contents lists available at ScienceDirect
Journal of Empirical Finance

journal homepage: www.elsevier.com/locate/jempfin
The real effects of nancial constraints: Evidence from a debt

subsidization program targeted at strategic rms
Yulia Davydova a, Vladimir Sokolov b,
a
b
CitiGroup Global Markets Prime Finance Risk, London, UK

International College of Economics and Finance, NRU Higher School of Economics, Moscow, Russia
a r t i c l e
i n f o
Article history:
Received 2 January 2014
Received in revised form 19 May 2014
Accepted 21 July 2014
Available online 26 July 2014
JEL classication:
G21
G28
G32
G38
Keywords:
Debt subsidy
Crisis
Strategic rms
Cash holdings
Banking development
Financial constraints
a b s t r a c t
We investigate the rationale and impact of the corporate debt subsidization program implemented
by the Russian government during the recent nancial crisis. Employing the difference-indifference approach, we show that the program did not have a signicant impact on capital
investments of subsidy recipients, contrary to its intentions. We also nd that a matched group
of non-recipients on average exhibited a higher degree of cash hoarding behavior than subsidyrecipients in the post-program period, which suggests that the program eased external nancial
constraints of recipient-rms. Consistent with the theory on precautionary cash savings by
rms, we further nd that rms non-recipients based in cities with low banking development
accumulated more cash holdings than recipients based in cities with a high banking development.
Overall, our ndings indicate that greater cash holdings are positively associated with the level of
nancial constraints of rms.
2014 Elsevier B.V. All rights reserved.
1. Introduction
The issue of how credit supply reduction impacts corporate policies of rms has received increasing academic attention in the recent period. For example, Campello et al. (2010) and Almeida et al. (2012) have demonstrated that a bank lending shock caused by a
subprime mortgage crisis has a pronounced real effect on nancially constrained rms. Following such shocks policy-makers often
respond with bailout programs aimed at relaxing the nancial constraints of rms considered crucial to the national economy.1
Existing academic work on such government-managed capital reallocations focuses on factors that make such programs benecial
to the real sector. Philippon and Schnabl (2013) study how government interventions can improve social welfare when banks suffer
from the debt overhang problem. Giannetti and Simonov (2013) provide microeconomic evidence that bank-recipients of
government-induced capital infusions increase lending to rms only if recapitalizations are sufciently large.
In this paper we study how the corporative policies of rms were impacted by the 2009 subsidization program implemented by
the Russian government following the banking crisis caused by the Lehman Brothers' collapse. The program was designed to subsidize
Corresponding author at: ICEF, NRU-Higher School of Economics, ul. Shabolovka 26, Room 3420, 119049 Moscow, Russia. Tel.: +7 495 772 95 90x26094.
E-mail addresses: yulia.davydova@citi.com (Y. Davydova), vsokolov@hse.ru (V. Sokolov).
1
Notable examples of such programs are the Troubled Asset Relief Program (TARP) that entailed recapitalization of too-big-to-fail banks, as well as the $25 billion
Advanced Technology Vehicles Manufacturing (ATVM) Loan program designed to support vehicle manufacturing in the U.S.
http://dx.doi.org/10.1016/j.jempn.2014.07.006
0927-5398/ 2014 Elsevier B.V. All rights reserved.
248
Y. Davydova, V. Sokolov / Journal of Empirical Finance 29 (2014) 247265
interest payments on bank loans taken by rms for investment purposes.2 This setting enables us to ask two related questions
regarding the investment and liquidity management policies of rms. First, did a program that was de jure designed to secure longterm investments by rms but de facto alleviated short-term crisis effects have the desired effect on subsidy recipients? Second, did
the impact of the program differ with respect to nancially constrained and unconstrained recipients?
We expect that, in the absence of an enforcement mechanism, rms did not signicantly increase capital investments upon
receiving interest rate subsidies from the government, although they may have changed their cash management policies. Further,
knowing that rms take into consideration both current and future funding needs when making policy decisions, we expect that
subsidy-receiving rms located in cities with a low level of banking development (nancially constrained rms) did not signicantly
change their corporate policies relative to non-recipients in similar locations. On the contrary, we expect that the impact of subsidies
on rms located in cities with a high availability of banking services (nancially unconstrained rms) was signicant compared to that
on non-recipients in similar locations.
The literature recognizes that observed and unobserved factors such as rms' political connections may simultaneously determine
the corporate policies of subsidy-recipients and government decisions to choose particular rms for subsidies (Faccio et al., 2006).
Because the program we study was targeted at so-called strategic rms with an average government ownership stake of 73%, the
heterogeneity of rms with respect to their political connections is greatly reduced.3 Focusing on strategic rms with signicant
government ownership and with equally high socio-economic value presents an opportunity to isolate rms that operate within
the same legal and institutional environment, have similar political connections, are at the same stage of the business cycle and differ
only in terms of observable characteristics that we can control. In view of this, the setup is ideal.
Studying the Russian experience is informative because we can exploit a substantial spatial diversity existing in the degree of banking development across the locations of strategic rms' and categorize rms into nancially constrained and unconstrained groups.4
Because the country's nancial system is bank-based, the variation in the supply of banking services across locations determines rms'
access to capital and is likely to be important in determining how they would respond to the subsidization of interest payments on
debt. This setting enables us to employ the difference-in-difference-in-difference (D-D-D) estimator (see Gruber, 1994; Imbens and
Wooldridge, 2007) and quantify the impact of the subsidization program across four groups of rms: (un)constrained (non)recipients.
Our analysis consists of several components. Using a sample of 357 strategic rms that were eligible for interest payment subsidies,
and controlling for industry afliation and the pre-crisis nancial and operating performance of rms, we rst show that the government was more likely to subsidize larger rms from smaller cities with low alternative employment opportunities. These ndings are
consistent with the government's public announcements on maintaining employment during the crisis period. This result is interesting in itself and contributes to the literature investigating government agendas for getting involved into the private sector (Frye and
Shleifer, 1997; Megginson and Netter, 2001; Shleifer and Vishny, 1994). We use this nding to validate our further analysis of the
post-subsidization performance of subsidy recipients.
In the following step, we match strategic rms that were subsidy-recipients (treated rms) with eligible strategic non-recipients
based on size, operating performance, industry afliation, and regional location using the propensity score matching (henceforth
PSM) algorithm. This method creates a sample of counterfactual control rms that are similar to recipients along the specied observable characteristics (see Roberts and Whited, 2012; Zhao, 2004). As a robustness check, we also isolate all strategic rms that were
nancially distressed in the pre-crisis period and test how the policies of distressed subsidy-recipients differed from those of
distressed non-recipients. Although isolating only distressed rms leaves the rms heterogeneous in other dimensions, this approach
narrows down the sample along the characteristics at which the subsidization program was directed.
We nd that, two years after completion of the government subsidization program, recipients had not signicantly changed their
investment and employment policies as could have been expected given program priorities. Our study also does not reveal any
signicant variation between nancially constrained and unconstrained rms with respect to long-term capital spending policies.
This nding is consistent with Giannetti and Simonov (2013) and Dasgupta et al. (2011) who document that while additional
short-term cash ows result in substantial cash savings by rms, the impact of such cash ows on long-term investment decisions
is marginal.
Our main ndings concern the cash management policies of rms. The theoretical literature predicts that, in periods of increased
uncertainty, rms experiencing constraints in external funding have incentives to increase cash holdings for precautionary reasons
(e.g., Almeida et al., 2004; Han and Qiu, 2007). We nd evidence consistent with this thesis: rms that did not receive subsidies
from the Russian government signicantly increased their cash holdings following the credit supply shock, while subsidy recipients
on average maintained the accumulation of cash at the pre-crisis level. Our difference-in-difference estimates suggest that cash
holdings of non-recipients grew by 6% relative to recipients for the PSM sample and by 10% for a sub-sample of nancially distressed
strategic rms. Thus, our results suggest that the program eased external constraints of treated rms.
Drawing on studies that examine the effect of spatial distribution of banking services on corporate polices (Benfratello et al., 2008;
Degryse and Ongena, 2005; Di Patti and Dell'Ariccia, 2004), we conjecture that rms that are located in cities with a high
2
The program used resources that were accumulated during the petro boom period in the Russian National Reserve Fund. Following public pressure this Fund was
untapped to help domestic rms suffering from the global nancial crisis.
3
A list of 1063 strategic rms was created by Presidential Decree No. 1009 adopted on August 4, 2004. According to the decree, the rms ascribed this status are those
that produce goods and services of foremost national priority. As strategic rms were granted this status long before the crisis and the subsequent subsidization program, possible concerns about sample selection bias are lowered.
4
We employ the BEPS II data set compiled by the EBRD, which reports geo coordinates (latitude and longitude) of all bank branches across the Russian Federation.
See Bircan and De Haas (2013) and Brown et al. (2013) for recent applications of the BEPS II data.
249
concentration of banks would consider themselves less constrained and the subsidization program would have a stronger effect on
them. Our D-D-D analysis shows exactly this: rms that received interest subsidies and are located in cities with a high supply of
banking services do not exhibit the cash hoarding behavior that matched non-recipients and recipients located in cities with a low
banking concentration do. We nd that depending on the analyzed subsample, the growth rate of cash holdings for nancially
unconstrained recipients was 1115% lower than that of the control group.
Our study contributes to several strands of academic literature. First, we add to an active literature on the value of cash for
nancially constrained rms (Denis and Sibilkov, 2010; Faulkender and Wang, 2006). Second, we contribute to the literature that
studies the interaction of local banking development and rms' corporate policies (see Benfratello et al., 2008; Bircan and De Haas,
2014; Di Patti and Dell'Ariccia, 2004; Jayarathe and Strahan, 1996). Finally we contribute to the literature that explores the economic
rationale of government involvement in the private economy (Frye and Shleifer, 1997; Lerner, 1999; Shleifer and Vishny, 1994). In
particular, our work is related to studies that examine government policies of bailing out banks during massive capital misallocations
in the banking sector (Duchin and Sosyura, 2011, 2012; Giannetti and Simonov, 2013).
The remainder of the paper is organized as follows: Section 2 sets out testable hypotheses; Section 3 describes the background of
the government subsidization program; Section 4 describes the data and provides summary statistics; Sections 5 and 6 report main
empirical results; and Section 7 concludes and discusses policy implications.
2. Hypotheses development
In this section, we review papers upon which we base our empirical investigation and testable hypotheses.
2.1. Government target selection decisions
What drives government assistance to the private sector? According to Frye and Shleifer (1997), the government's interaction
with rms in emerging market economies can be described according to either the grabbing-hand or helping-hand models.
Within the helping-hand model, one can distinguish between cases when government interventions are designed to address market
failures due to unforeseen negative shocks and cases when the helping-hand is extended to a priori distressed rms. In the rst case,
one expects that the government would provide subsidies to well-performing rms hit by crisis. In the latter case, an exogenous event
such as the collapse of the Lehman Brothers is benecial to a priori distressed rms because it triggers government help that they
would not otherwise receive. This motivates our rst hypothesis.
Hypothesis 1. Negative performance of strategic rm in the pre-crisis period is positively related to the likelihood of a rm being
chosen for subsidization.
When public funds are dispersed, non-economic motives cannot be neglected. We argue that by focusing on strategic rms
with signicant government ownership, we reduce the likelihood that our analysis would be affected by variations in the
political connectedness of rms. However, political factors are still relevant for our study. In an environment where the
distribution of government subsidies is implemented through a top-down policy, as was the case in Russia, the desire of highlevel political executives to retain popular support and be re-elected could be a signicant factor inuencing such programs.
By helping distressed strategic rms to stay aoat, politicians may seek to prevent the possibility of social unrest caused by
massive bankruptcies (Shleifer and Vishny, 1994).
The possible impact of political connections on the allocation of government subsidies is another issue worth consideration. Due to
the top-down design of the subsidization program that we study, we cannot exclude that subsidization decisions were inuenced by
political lobbying targeting the government ofcials in charge of this program. We expect that such lobbying could have been
conducted by parliamentary deputies from the ruling United Russia party, who are likely to have connections among government
ofcials, on behalf of rms from the regions they represent. We focus on the deputies in the lower house of the parliament (the
Duma)5 and arrive at the following hypothesis:
Hypothesis 2. The relative level of ruling party representation in the Duma for a particular region is positively associated with the
probability that a strategic rm located in this region will receive subsidies, as well as with the size of the subsidy.
The political cost of the bankruptcy of a strategic rm can be expected to depend on the rm's location and the size of the city
where it is based. The phenomenon of one-company towns is particularly pervasive in Russia due to the Soviet command economy
legacy (see Commander et al., 2011). For example, a May 2009 protest in the small town of Pikalevo against the shutdown of a local
factory received a lot of media coverage and resulted in the direct involvement of the Prime Minister (see Galpin, 2009, BBC News).
Since small cities have lower alternative employment opportunities and a higher potential for social discontent in the case of the
bankruptcy of a strategic rm, it can be expected that rms located in such cities are more likely to receive government subsidies.
Following this logic, we also expect that rms located in remote cities are more likely to be help-recipients. Aside from the size and
5
During the period that is being studied (the 2007 Duma elections), all Duma members were elected on party lists in a single nationwide constituency under proportional representation. These party lists contain both nation-wide top candidates and regional-level candidates, categorized in groups dened by the Central Electoral
Commission that correspond to territories with a certain share of votes (including regions, parts of regions and several regions joined together). The party leadership
decides on the inclusion of candidates from different regional groups on the party lists, while voters vote for party lists as a whole. According to ofcial results, the ruling
United Russia party won 64% of the votes. The Communist Party, the Liberal Democratic Party and Just Russia also gained seats in the parliament.
250
remoteness of the city, another measure of the alternative employment opportunities of workers is the Herndahl index (HHI) of
employment concentration for each city. This leads us to the third hypothesis.
Hypothesis 3. The size of the city where a strategic rm is located is negatively related to the likelihood of a rm being chosen for
subsidization. The city's distance from the capital and the level of employment concentration in the city captured by the HHI index
are positively related to such likelihood.
2.2. The impact of the program on rms' corporate policies

When addressing the main questions of our study concerning the consequences of the government assistance program, we take
our starting point in the predictions of a model developed by Almeida et al. (2011) on the relationship between real investments
and intertemporal nancing frictions. The authors demonstrate that when credit constraints are likely to bind in future, rms prefer
investments with shorter payback periods regardless of current nancing conditions. Due to high uncertainty regarding future cashows following the severe economy-wide shock caused by the global nancial crisis, as well as the transitory nature of the
subsidization program we do not expect any discernible patterns in capital investments across subsidy-recipients and non-recipients.
Hypothesis 4. Interest payment subsidy recipients do not signicantly change their long-term capital spending relative to matched
non-recipients.
A model developed by Han and Qiu (2007) provides a theoretical link between rms' cash holdings and external nancial constraints. These authors show that it is optimal for a nancially constrained rm to save cash for precautionary reasons so that these
resources can be utilized for future investment opportunities. At the same time, a nancially unconstrained rm has no precautionary
motives for increasing its cash holdings according to their model. Since the subsidization of interest payments in Russia increased free
cash-ow of rm-recipients, we expect that external constraints facing them were relaxed. This is the basis for our next hypothesis.
Hypothesis 5. Subsidy recipients maintain cash holdings at the pre-crisis level, while matched non-recipients exhibit a relatively
higher accumulation of cash.
All models on cash hoarding behavior of rms (see Almeida et al., 2004, 2011; Han and Qiu, 2007) stress that such effect depends
on the assumption by rms that they will be nancially constrained in the future. Our data allows us to directly test this prediction. We
posit that during a period of high uncertainty, the policy of interest payment subsidization is viewed by rms as transitory and rms
located in cities with a low bank presence are bound by their future external nancial constraints. On the contrary, in areas with a high
level of banking development and possibly stronger rmbank relationships, the subsidization policy results in non-hoarding
behavior among treated rms.
Hypothesis 6. Following credit-supply shock non-recipients of subsidies located in cities with low banking development increase
cash holdings relative to recipients located in cities with high banking development.
3. Background of the debt subsidization program

3.1. Macroeconomic environment
The collapse of Lehman Brothers has been used as an identifying event in a number of recent studies (see Almeida et al., 2012;
Duchin et al., 2010). For the Russian nancial system this event represented a sudden-stop in foreign borrowing, which triggered
an economy-wide credit supply shock of exogenous nature.6 A precipitous increase in the cost of foreign borrowing for Russian companies reected capital market concerns about the country's banking system, which heavily relied on foreign funding at that time.
The sudden-stop of foreign borrowing resulted in capital misallocations that led to a signicant decline of the manufacturing
sector. As can be seen from Fig. 1, in the last quarter of 2008, the seasonally adjusted aggregate manufacturing index fell by 15%.
Following these developments, the Russian government untapped resources accumulated during the petro boom period in the National
Reserve Fund and initiated a massive anti-crisis program. Fig. 2 illustrates the depletion of the reserve fund following the Lehman event.
This anti-crisis program was implemented over the course of 2009 and allocated assistance to nancial institutions, vulnerable groups
of the population and to non-bank rms. Information about the identity of recipients, and the size and type of aid can be found in the
report Anti-crisis Measures for 2009 published by the Russian Ministry of Economic Development in 2010.7
6
From January 2008 till October 2008 the CDS spreads on 5-year Russian sovereign debt ranged between 85 and 117 basis points. Following the Lehman's bankruptcy
by the end of October 2008 the CDS spread on Russian sovereign debt went up to 1100 basis points. Over the course of 2009 the uncertainty about scal position of the
Russian government subsided and the spread gradually declined to 150 basis points.
7
The government recapitalized large banks and provided uncollateralized loans to banks with investment grade credit ratings (See Sokolov (2012)). National champions such as railroads, air carriers and auto manufactures received loan guarantees on new loans at state-owned banks such as Sberbank, VTB, VEB, Gazprombank and
Rosselhozbank.
251
Fig. 1. Dynamics of the seasonally adjusted manufacturing index.

Source: Rosstat.
3.2. Subsidization of the strategic rms program

Our study focuses on the outcomes of one of the anti-crisis programs completed by the Russian government subsidization of
strategic manufacturing rms' interest payments on bank loans. The program was launched by the Ministry of Industry and Trade
on December 31, 2008 and allocated funds equivalent to 300 mln. USD over the course of 2009. According to the program guidelines,
an eligible strategic rm could receive subsidies on up to two thirds of its interest rate costs during the period for which the subsidy
was provided. The program allowed subsidization of bank loans taken in Russian or foreign banks for the purpose of investing into
new equipment.
4. Summary statistics and data description
4.1. Sample formation
We start with a sample of strategic rms that were assigned this status by Presidential Decree No. 1009 adopted on August 4,
2004.8 Over the pre-crisis period, the government repeatedly added and excluded rms from this list. Firms were dropped from
the list when they were merged into state holding companies. We have traced these changes and use the latest pre-crisis list of
rms that have been ofcially designated as strategic rms.
We exclude all rms that according to ISIC Rev.4 industry codes do not belong to the manufacturing sector and thus were ineligible
under the subsidization program.9 We further exclude all holding companies from the list. We match all the remaining eligible
strategic rms with the rm-level data obtained from the Ruslana database provided by the Bureau van Dijk. The data includes
20062011 annual unconsolidated balance sheets, and income and cash ow statements of rms registered in the Russian Federation.
We keep all rms that have at least one non-missing observation of nancial variables in the pre-crisis and post-subsidization periods.
Being left with a sample of 357 strategic rms eligible for government aid, we match it with a list of subsidy recipients published in
the government report on Anti-crisis Measures for 2009 and identify 65 rms that received interest rate subsidies. Table 1 reports
the distribution of the rms' industry afliations in our sample by the 3-digit ISIC primary industry codes. As can be seen from the
fourth and fth columns of the table, industries that were strongly represented in the program include: industries manufacturing
ships and boats, defense industries and air and spacecraft industries. Out of all rms that belong to the ship building industry 75% received interest subsidies, while 31% of the defense industry and 28% of the aircraft building rms were recipients. The average interest
payment subsidies accounted for 9.22% of the rm-recipients' pre-crisis assets, which suggests that the government aid was quite
sizable and likely to substantially relieve the nancial constraints of rms.
In order to gauge the degree of banking development in the cities where strategic rms are located, we employ the BEPS II data set
compiled by the EBRD, which contains geo coordinates of all bank branches across the Russian Federation. Table 2 provides summary
statistics of the distribution of the banking branches in 113 cities where strategic rms in our sample are located. Fig. 3 illustrates the
distribution of these cities across the country.
Since the government subsidization program was aimed at alleviating the strategic rms' nancial constraints, we dene a subsample of nancially distressed rms that we study in parallel fashion throughout our paper. Following the literature (Andrade and
Kaplan, 1998; Asquith et al., 1994), we classify a rm as being nancially distressed if it had a negative value of the interest rate
8
According to the decree, the 1063 rms beneting from this status are those that produce goods and services of rst national priority.
We exclude rms belonging to the following industrial sectors: agricultural and food products (011031); production of electricity, gas and oil (351353);
construction (410439); transportation and storage (491532); and media and entertainment (900932).
9
252
Naonal Reserve Fund of the Russian

Fedaraon
USD. Blns.
150
100
50
01.02.2008
01.04.2008
01.06.2008
01.08.2008
01.10.2008
01.12.2008
01.02.2009
01.04.2009
01.06.2009
01.08.2009
01.10.2009
01.12.2009
01.02.2010
01.04.2010
01.06.2010
01.08.2010
01.10.2010
01.12.2010
Fig. 2. Dynamics of the National Reserve Fund.

Source: Ministry of Finance of the Russian Federation.
coverage (EBITinterest payments) during any two consecutive years in the 20062008 pre-subsidization period. According to this
criterion, 43 strategic rms are dened as nancially distressed with only 22 of them being subsidy recipients.
4.2. Description of variables

A detailed description of all variables used in our analysis can be found in Appendix A. These variables can be categorized as
follows: nancial characteristics of strategic rms, geographic and socio-economic characteristics of cities where strategic rms are
located, and characteristics of these cities in terms of their banking development.
4.2.1. Financial characteristics of rms

In evaluating the rationale of the subsidization program, we follow prior studies on interaction of the state with the private sector
(Boubakri et al., 2009; Megginson et al., 2004) and employ a number of control variables in our analysis, as described in the following.
Total revenue captures the rm's size. Employment reects the total number of employees. A measure of asset, Tangibility is tangible
xed assets divided by total assets. Protability is dened as the operating income before interest and taxes scaled by total assets.
The rm's Debt-to-assets is dened as the ratio of total debt to total assets. Cash holdings are cash and cash equivalents divided by
total assets. Assets turnover measures a rm's efciency in using its assets in generating revenue and is captured by the ratio of
total revenue to total assets.
In a subsequent difference-in-difference analysis of the subsidization program's impact on strategic rms, we follow Giannetti and
Simonov (2013) and concentrate on the growth rates during the pre- and post-program periods of the following variables: change of
cash holdings scaled by initial total assets, change of total debt scaled by initial total assets, growth of employment, and change in net
capital expenditures scaled by initial total assets.
4.2.2. Geographic, socio-economic and political characteristics of cities

We next turn to characteristics of the cities where strategic rms are located. Previous studies like Cull and Xu (2005) on China and
Bruno et al. (2013) on Russia show that variation in regional characteristics has an impact on rms' investment and market entry
decisions. Motivated by the work of Commander et al. (2011) and ofcial government announcements around the program
completion, we conjecture that city size, distance to the federal capital and average year temperature proxy for alternative employment
opportunities in cities where strategic rms are located and should be related to the government's selection process of subsidy
recipients. We also calculate the local Herndahl index (HHI) of employment concentration for each city by collecting data on
employment by all rms located in the cities of our sample at the time of the subsidization program.
Further, following Caselli and Michaels (2013), we proxy regional infrastructure with highway length per 1000 km2 and use this
proxy as a control variable in our regressions. Compared to other region-specic variables, this proxy is time-invariant and weakly
exogenous with respect to subsidization program outcomes.
In order to test the hypothesis that government subsidization decisions were inuenced by political lobbying conducted by ruling
party deputies advocating for rms based in their regions, we have constructed the following variable: the number of Duma deputies
from a particular region who represent the ruling United Russia party divided by the region's population.10
10
We have experimented with several measures of political inuence, such as the proportion of ruling party deputies among the total number of deputies
representing a particular region and the concentration of opposition party deputies from a particular region. None of them turned out to be signicant.
253
Table 1
Distribution of strategic companies and subsidy recipients across industries. The third column reports a number of strategic rms that belong to each industry. Column 4
reports a number of subsidy recipients within each industry respectively. Industries are assigned by the ISIC primary codes. The Defense industry has been manually
assigned by checking companies' web-sites and ofcial lings. Column 5 reports an average of Subs.-to-assets across rm recipients in a given industry. The size of interest payment subsidies for each strategic rm are taken from the Government Report on Anti-crisis Measures for 2009 published by the Ministry of Economic Development. Subsidies were distributed in 2009. Pre-crisis assets are determined: Pre-crisis Assets = (Assets2006 + Assets2007) / 2.
Industrial sector
ISIC Rev. 4 codes
Total number of Number of rms receiving Interest payment subsidies/

strategic rms
interest payment subsidies pre-crisis assets (%)
(1)
(2)
(3)
(4)
Chemicals; rubber and plastic; Basic and fabricated metals 201203; 221222;
28
241259
Manufacture of communication measuring, testing
263; 265; 271; 273 108
equipment; electric motors, generators, transformers
Manufacture of machinery and motor vehicles
281282; 291293
16
Manufacture of ships, boats and railway locomotives
301302
12
Manufacture of air and spacecraft
303
43
Scientic research and development
721
73
Defense
202; 261263; 301;
77
304; 721
357
(5)
22.91
20.32
4
9
12
4
24
11.45
12.13
3.68
5.92
5.10
65
9.22
4.2.3. Banking development characteristics

In order to gauge the level of banking development in cities where strategic rms are located we calculate two variables. The rst
one is branch density calculated as the total number of bank branches per capita in the city. This variable has been used by Benfratello
et al. (2008) and by Di Patti and Dell'Ariccia (2004) as a measure of the level of development of local credit market of Italian rms.
Holding other things constant, a rm located in a city with higher branch density should be less nancially constrained.
Another variable that we use is proportion of regional bank branches in the city. We calculate it as the number of branches of banks
founded at the regional level in relation to the overall number of banks in the city. This variable is motivated by studies of Degryse and
Ongena (2005) on Belgium and Bharath et al. (2007) on the US, both of which provide evidence that regional banks have stronger
lending relationships with local rms than nation-wide banks. In case of a credit supply shock, one can expect that rms with strong
Table 2
Descriptive statistics of the pre-crisis strategic rms' characteristics. The left panel reports summary statistics for the variables that characterize all strategic rms eligible
for the government subsidization program. The right panel reports statistics for a subsample of strategic rms that were nancially distressed before the subsidization
program. Financially distressed rms are dened as those that had negative EBITinterest payments in any two consecutive years before the subsidization program
(20062008). All rm characteristics are calculated as an average over 20062007 annual values of corresponding ratios. Sub.-to-assets is calculated for recipients only,
where for each rm we take the amount of help received in 2009 with respect to average 20062007 assets. When constructing the banking development variables we
match BEPS II data on locations of all banks' branches with the locations of all strategic rms. The table reports summary statistics of the concentrations of bank branches
and the proportion of regional banks in the cities where strategic rms are located.
Financially distressed rms sub-sample
Full sample
Mean
Firm variables
Subsidy/assets
Revenue, mln. rub.
Total assets
Num. employees
Tangibility
Cash-to-assets
Debt-to-assets
Protability
Assets turnover
Interest coverage
Capex-to-assets
State share
StDev
Min
P50
Max
0.092
0.145
0.001
0.020
0.474
1539
3675
68.36
469.3
33,795
2748
8567
30.45
566.2
93,148
1165
2573
10
499
30,465
0.281
0.174
0.007
0.241
0.778
0.078
0.096
0
0.039
0.532
0.115
0.165
0
0.032
0.738
0.079
0.075
0.115
0.069
0.344
1.044
0.649
0.083
0.888
3.639
154.9
688.1
76.50
2.497
8153
0.029
0.037
0
0.016
0.206
0.727
0.315
0.001
1.00
1.00
Location characteristics
City population
443
1093
Dist. to capital
1041
1419
HHI employment
0.062
0.092
Ruling party conctr.
2.137
0.507
Average temp. C
4.73
2.53
Highway per area
248
236
City's latitude
54.57
3.939
City's longitude
50.71
23.40
Bank branch density
0.408
0.679
Prop. of reg. bank branches
0.127
0.131
7.50
175.4
10,509
0.00
610
6442
0.00
0.020
0.468
0.92
2.10
3.721
5.3
4.9
13.4
6
175
1239
42.98
55.60
64.56
20.50
41.91
137.0
0.070
0.324
7.278
0
0.099
0.894
Obs Mean
StDev
Min
P50
Max
Obs
65
0.144
0.169
0.002
0.034
0.474
357 2986
5919.3
79.88
424.6
33,795.5
357 8236
17,672
163
1524
93,148
356 3079
6193
16
689
30,465
357
0.287
0.201
0.031
0.243
0.720
357
0.029
0.051
0
0.011
0.247
357
0.264
0.240
0
0.174
0.738
357
0.020
0.043
0.115
0.015
0.069
357
0.573
0.349
0.083
0.528
1.527
357
4.655
19.43
76.50
0.090
61.97
350
0.017
0.025
0.000
0.009
0.109
357
0.70
0.315
0.001
80.00
1.00
22
43
43
42
43
43
43
43
43
43
43
43
113
358
113 1416
113
0.089
113
1.88
113
4.252
113
235
113
55.2
113
56.21
113
0.682
113
0.117
19
19
19
19
19
19
19
19
19
19
374
1875
0.134
0.570
1.999
297
4.99
30.79
1.6
0.109
10.5
24
0.002
0.921
0.80
10
44.16
31.03
0.095
0
190.7
839
0.014
1.932
4.4
128
55.72
44.53
0.341
0.064
1135
6404
0.467
3.454
8.10
1239
64.56
135.1
7.28
0.298
254
Fig. 3. The geographical distribution of the location of strategic rms across the Russian Federation.
Source: Google maps, Ministry of Economic Development of the Russian Federation.
255
lending relationships with local banks should be less nancially constrained. One would expect the dependence of Russian rms on
local banks to be even higher. It is well documented that local business groups created so-called captive local banks, which they used
for conducting their nancial transactions, during the transition period in Russian. This partially explains the existence of as many as
around 1000 banks in Russia.
We denote a rm as being nancially unconstrained if the banking concentration or the proportion of regional bank branches in
the city where it is located is above the sample medians of these two proxies.
4.3. Summary statistics and univariate analysis

In order to investigate the rationale and consequences of the government subsidization program, we need to measure rms'
pre-crisis characteristics. The rm-specic variables are calculated as averages of the key nancial ratios over 20062007.
Focusing on a period of two years prior to the program's completion ensures that rms' characteristics are pre-determined.
We also report statistics of time-invariant geographic and socio-economic variables for the cities where strategic rms are located.
We winsorized all continuous rm-specic variables at the 2.5% and 97.5% levels.
The left panel of Table 2 presents summary statistics of the full sample of strategic rms, the right panel of the same table focuses
on a sub-sample of nancially distressed rms. As expected nancially distressed rms are on average more indebted and exhibit less
protability and lower asset turnover than non-distressed rms. One can also observe that distressed rms are larger by all size
characteristics such as: revenue, total assets and employment.
Table 3 helps to identify along which characteristics recipients and non-recipients differ for both sub-samples. Let us rst focus on
the results for the full sample of eligible strategic rms, which are reported in the left panel of the table. Mean difference t-tests and
median difference z-tests uniformly reject the null that the key nancial ratios of recipients and non-recipients are the same during
the pre-crisis period. On average, rms selected for the subsidization program were about four times larger, two times more indebted,
held two times less cash and were two times less protable than non-recipients. At the same time, the difference in government ownership stake is insignicant between the two groups of rms.
Univariate tests on rms' location characteristics for the full sample of strategic rms suggest that rm-recipients were located in
smaller cities, further away from the capital and with lower median year temperatures than those where non-recipients were based.
The density of highways per square kilometer suggests that rm-recipients came from regions with less developed infrastructure. In
Table 3
Univariate tests for pre-crisis rms' characteristics between interest subsidy recipients and non-recipients. The left panel of the table reports sample differences
between subsidy recipients and non-recipients for all strategic rms eligible for subsidies. The right panel of the table reports differences between recipients and
non-recipients for a subsample of strategic rms that were nancially distressed before the subsidization program. Financially distressed rms are dened as those that
had negative EBITinterest payments in any two consecutive years before the subsidization program (20062008). All rm characteristics are calculated as an average
over 20062007 annual values of corresponding ratios. The difference tests report t-test (Wilcoxon rank sum z tests) values for the difference in means (medians)
between non-recipients and recipients.
Financially distressed rms sub-sample
Full sample
Firm variables
Revenue, mln. RUB
Tangibility
Cash-to-assets
Debt-to-assets
Protability
Assets turnover
State ownership
share
Non-recipients
(292 rms)
Recipients
(65 rms)
Mean
Mean
957.8
0.295
0.086
0.090
0.088
1.135
0.73
Median
Median
Difference tests
(non-recip. Recip.)
Non-recipients
(21 rms)
Recipients
(22 rms)
t-Test
Mean
Mean
Wilcoxon
z-test
Median
Difference tests
(non-recip. recip.)
Median
t-Test
Wilcoxon
z-test
420.3
4148.1
1293.5
3.853*** 4.75***
2841.9
296.7
3124.1
1807.8
0.152 1.433
0.275
0.219
0.181
3.378***
3.422***
0.325
0.356
0.250
0.199
1.228
1.118
0.048
0.041
0.019
4.773***
4.168***
0.03
0.013
0.029
0.006
0.033
0.316
0.017
0.226
0.164 5.221*** 5.836***
0.228
0.104
0.298
0.356 0.955 0.839
0.075
0.039
0.038
5.329***
4.661***
0.025 0.032
0.015
0.000 0.711 1.093
0.984
0.632
0.536
8.001***
7.142***
0.665
0.634
0.485
0.450
1.714*
1.531
1.00
0.72
1.00
0.229
0.449
0.771
1.00
0.642
0.65
1.355
1.512
Location characteristics
City population
4185
1280
3024
1094
2.076**
Distance to capital
593
374
1135
634
2.414**
0.029
0.161
0.032
0.010 0.563
HHI employment
Ruling party
1.897
1.839
1.914
1.839 0.323
concert.
Ave. year
5.164
5.8
4.821
5.3
1.534
temperature
High way per area
0.303
0.197
0.261
0.165
1.359
Prop. of regional
0.389
0.260
0.313
0.209
1.754*
bank branches
Bank branch
0.886
0.465
0.817
0.386
0.537
density
Notes: *, **, and *** denote signicance at the 10%, 5% and 1% respectively.
1.253
2.64***
0.861
0.599
3275
611
1497
821
634
1633
0.053
0.009
0.046
1.784
1.915
1.952
605
1.637
703
1.526
0.009
0.241
1.899 1.081
0.986
1.011
0.037
1.036
1.963**
4.505
5.8
4.7
4.9
0.375
0.201
2.048**
1.345
0.243
0.335
0.128
0.259
0.249
0.199
0.163 0.079
0.184
1.611
0.160
1.158
1.461
0.738
0.379
0.867
0.385 0.376
0.353
256
terms of local banking development of cities, the two-sample t-tests (z-tests) indicate that the subsidy-recipients and non-recipients
were not different.
The right panel of Table 3 presents the comparison of subsidy-recipients and non-recipients for a sub-sample of strategic rms that
were nancially distressed in the pre-crisis period. The two-sample tests reveal no signicant difference between distressed recipients
and distressed non-recipients neither along rm-specic nor location characteristics.
Overall, these univariate comparisons suggest that an average subsidy recipient is signicantly different from an average nonrecipient in the full sample of eligible strategic rms. However, characteristics for nancially distressed groups from both groups
are similar and further multivariate analysis is needed to determine the government's agenda when selecting rms for its subsidization program.
5. The determinants of the government subsidization program

In this section we further investigate the suggestive results of the univariate analysis by conducting multivariate regression
modeling of the government's rationale in selecting strategic rms for subsidies. We employ probit and tobit models where the probit
specication captures incidences of strategic rms being subsidy-recipients, while the tobit specication uses Subsidies-to-assets ratio
as a dependent variable. We calculate this ratio by scaling the size of subsidies that each rm received in 2009 with average rm assets
in the 20062007 period.
Using mortgage sub-prime crisis as an identifying event, Duchin et al. (2010) argue that rms' pre-crisis decisions on balance sheet
structure are uncorrelated with crisis-period shocks. Following this identication strategy, we calculate rms' pre-crisis characteristics as averages of nancial variables over the 20062007 period and regress outcomes of the 2009 subsidization program on these
rm-specic and location characteristics. This setting allows us to test Hypotheses 13 on the government's agenda in distribution
of subsidies to rms.
Table 4
Firm characteristics and debt subsidization program. The left panel of the table reports estimated marginal effects of lagged rm characteristics at their means on the
probability of receiving interest rate subsidy in 2009. The right panel reports marginal effects of each explanatory variable on the size of subsidies, given that the rm has
been a subsidy-recipient (i.e., has not been censored). Subs./assets ratio is calculated as a ruble value of the subsidy received by the rm in 2009 to the rm's average
value of the total assets over 20062007. All rm-specic explanatory variables (except dummies) are calculated as an average over 20062007 annual values of corresponding ratios. The estimates reported in columns 2 and 5 exclude strategic rms that are located in Moscow. The estimates reported in columns 3 and 6 focus on
nancially distressed rms which are dened as those that had negative EBITinterest payments in any two consecutive years before the subsidization program (2006
2008).
Tobit (marginal effects E (Subs|Subs N 0))
Probit (marginal effects)
Firm size (log revenue)

Tangibility
Cash-to-assets
Debt-to-assets
Protability
Assets turnover
State ownership share
City population
Dist. to capital
HHI Employment
Average year temperature
Ruling party region. concentr. of deputies
Highway length per area
Industry effects
N. of obs.
Left censored
Pseudo R2
Full sample
Exclude Moscow
Financially distressed
Full sample
Exclude Moscow
(1)
(2)
(3)
(4)
(5)
Financially distressed
(6)
0.044***
(3.38)
0.088
(0.93)
0.078
(0.44)
0.174**
(2.49)
0.426
(1.06)
0.150***
(3.20)
0.001*
(1.65)
0.001
(0.59)
0.001***
(3.05)
0.222
(1.03)
0.009
(1.23)
0.042
(0.97)
0.001
(0.63)
Yes
357
0.037***
(2.58)
0.144
(1.18)
0.007
(0.03)
0.230*
(1.88)
1.299***
(3.60)
0.112**
(2.07)
0.001*
(1.83)
0.001**
(2.12)
0.001**
(2.00)
0.306
(1.13)
0.013
(1.45)
0.062
(1.31)
0.001
(1.16)
Yes
254
0.498**
(2.55)
9.522***
(2.68)
0.907
(0.17)
2.053
(1.61)
4.222
(1.13)
3.239***
(3.02)
0.36
(0.66)
0.001**
(2.55)
0.001
(0.57)
0.316
0.339
0.686
0.001
(0.15)
0.22
(1.13)
0.024
(0.55)
0.018
(1.30)
0.271***
(4.03)
0.010
(1.06)
0.001**
(2.00)
0.001**
(2.20)
0.001*
(1.91)
0.009
(0.20)
0.002
(1.36)
0.013*
(1.65)
0.001
(1.41)
Yes
254
202
0.530
0.001
(0.20)
0.146**
(2.24)
0.088
(0.62)
0.012
(0.24)
0.567**
(2.44)
0.019
(0.52)
0.051*
(1.78)
0.001
(1.45)
0.001
(0.94)
Yes
43
0.002
(1.09)
0.012
(0.62)
0.020
(0.66)
0.018**
(2.12)
0.137*
(1.74)
0.019**
(2.27)
0.001*
(1.65)
0.001
(0.24)
0.001***
(3.69)
0.011
(0.23)
0.002
(1.42)
0.01
(1.32)
0.001
(0.06)
Yes
357
292
0.469
Yes
43
21
0.537
Notes: *** p b 0.01, ** p b 0.05, * p b 0.1. t-Statistics in parentheses. In order to control for the intra-regional correlation we cluster standard errors at the regional level.
257
5.1. Probit results

Results of the multivariate analysis are reported in Table 4 and reveal that strategic rms receiving subsidies are large, have high
leverage and poor performance, and are located in smaller and more remote cities. Columns 1 to 3 of Table 4 report marginal
probability effects from probit model estimation relating the probability of an average rm being chosen by the government for
subsidization to various rm characteristics.11 Column 1 shows the results for the full sample of eligible strategic rms, column 2
shows the results for the sample of all non-Moscow rms and nally columns 3 shows the results for the sub-sample of nancially
distressed rms.
We nd that the rm's size measured by total revenue has a strong positive effect on the probability of it receiving subsidies for all
three examined samples. The results are economically signicant. The coefcient estimate for the full sample of strategic rms suggests that a two-standard-deviation increase in a natural logarithm of average pre-crisis revenues leads to a (2.496 4.4%=)
10.98% increase in the probability of a rm being a subsidy recipient. Two other statistically signicant coefcients suggest that for
an average strategic rm a one percent increase in the rm's leverage leads to a 17.4% increase in the probability of being selected
for subsidization. At the same time, a one percent increase in asset turnover of an average strategic rm leads to a 15% decline of
this probability which means that rms that exhibited inferior operating performance during the pre-crisis period were more likely
to get help, other things being equal.
As visible from column 2 of the table, all results are robust to the exclusion of strategic rms located in Moscow. As argued in the
Introduction, an important factor that might affect the government's agenda in distributing aid to strategic rms is related to the desire to keep these rms aoat in order to prevent social unrest in case of massive bankruptcies. As expected, rms located in smaller
and remote towns with low alternative employment opportunities are more likely to be interest rate subsidy recipients. The characteristics of cities where strategic rms are located indicate that an increase in the city's population by 100 thousand inhabitants is associated with a 10% drop in the probability of a rm from such a city to be selected for government subsidies. An increase in the
distance between the city where an average strategic rm is located and the capital by 100 km increases this probability by 10%.
The HHI employment index turns out to be insignicant, which suggests that this variable is a crude measure of the city's social value.
An interesting picture emerges from the sub-sample of nancially distressed rms reported in column 3. Contrary to the univariate
results, the estimation of the multivariate probit model suggests that pre-determined characteristics of the distressed rms' such as:
the rm's size, tangibility, asset turnover and the size of the city where a rm is located are signicantly related to the probability of a
rm being selected for the subsidization program. It is important to note that each reported marginal effect represents a slope of a
non-linear predication function evaluated at the sample mean of the regressors. In cases when the slope of the prediction curve changes quickly the approximation provided by the derivative evaluated at a point becomes less precise. This explains estimated marginal
effects larger than unity for the distressed rm sub-sample.
5.2. Tobit results
As the next step, we undertake a multinomial tobit analysis in order to examine the intensive margin of the subsidization program.
We utilize the Subsidies-to-assets ratio as a dependent variable and present the results in columns 46 of Table 4. When calculating the
marginal effects, we use a specication that takes into account the censored nature of the dependent variable. The table reports marginal effects of each explanatory variable on the expected size of subsidies, given that a rm has not been censored (i.e., has received a
subsidy) across the three examined sub-samples. Our results indicate that the expected size of a subsidy to program beneciaries is
signicantly positively related to the rm's pre-crisis leverage and negatively related to its protability. Consistent with the evidence
from the probit model, the size of the city where an average rm is located is negatively related to the size of the subsidies it received.
The estimates of the coefcient on the variable measuring the relative level of ruling party representation in the Duma for different
regions are positive. For a sample where Moscow-based rms are excluded (reported in column 5), the coefcient is marginally signicant. This suggests that the level of representation of the region where a strategic rm is located is positively associated with the
size of the subsidies allocated to such a rm.
In summary, our results cannot reject Hypothesis 1 that negative performance by strategic rms during the pre-crisis period is signicantly associated with the likelihood that they were selected for government subsidies. Similarly our results cannot reject
Hypotheses 2 and 3, according to which the level of ruling party representation in the parliament and the characteristics of the cities
where rms are located are signicantly associated with the likelihood that rms were granted subsidies.
6. The impact of the subsidization program on rms' corporate policies
6.1. Propensity score matching of strategic subsidy recipients and non-recipients
Empirical research that examines the impact of credit-supply shocks on rms (see Lemmon and Roberts, 2010), as well as studies
that evaluate the consequences of government subsidies on rms (see Lerner, 1999) recognize that an average treated rm may be
signicantly different from an average non-treated rm along various characteristics. In accordance with this literature, an unbiased
estimation of the post-crisis response of rms to the subsidization program requires that the so-called parallel trends assumption is
11
We report marginal probability effects of the individual regressors evaluated at sample means of all independent variables.
258
satised. Under this assumption, trends related to rms' corporate policies must not be signicantly different across treated and nontreated rms during the pre-crisis period.
In our setting, the strategic rms represent a homogeneous group that was pre-selected by the government long before the crisis
and has similarly high state ownership and social value. This setting greatly reduces concerns about unobserved factors that may
confound the inference. At the same time, our analysis of the program rationale in the previous section shows that even if strategic
rms are arguably similar along many important dimensions, the selection of subsidy-recipients by the government was not random.
In order to address confounding factors in the comparison of treated and non-treated rms, we employ the propensity score method
(henceforth PSM) that creates a matched sample of treated and control rms with similar propensity scores. This method allows
controlling for potential differences in the observable characteristics of rms and eliminating bias in the estimation of the subsidization program outcomes.
In implementing the propensity score matching algorithm we follow Zhao (2004), Lemmon and Roberts (2010), and Roberts and
Whited (2012) and rst estimate the probit regression on our full sample of strategic rms and obtain the recipients' (treated rms)
and non-recipients' (non-treated rms) propensity to be (or likelihood of being) a subsidy recipient. We match non-treated rms
with treated ones on the basis of nancial characteristics of rms that were found to be signicant in our previous analysis, rms'
industry afliation and the region where the rms are located. The rst column of Table 5, Panel A presents the probit model
estimation results. Using the propensity scores from this model we perform a one-to-one nearest-neighbor matching without
replacement. That is for each subsidy recipient we identify the non-recipient with the closest propensity score. We impose a common
Table 5
Propensity score matching diagnostics. The sample begins with 375 strategic rms eligible for interest rate subsidies. The treatment group includes all rms that
received interest rate subsidies in 2009. The control group is matched by the propensity score method. We use one-to-one nearest neighbor without replacement
specication and impose a common support restriction which drops treatment observations whose propensity score is higher than the maximum or less than the
minimum propensity score of the controls. Panel A reports coefcient estimates from the probit model used in estimating the propensity scores for the treatment
and control groups. The pre-match column reports parameter estimates for the entire sample of strategic rms (this specication generates propensity scores for
matching). The post-match column reports parameter estimates for the subsample of matched treatment and control observations after matching. Panel B reports
pairwise comparison of variables on which matching is done for samples before and after matching.
Panel A. Probit regression results

Protability
Assets turnover
Debt-to-assets
Cash-to-assets
Region index
Industry index
Constant
N. of obs.
Pseudo R2
Pre-match
Post-match
0.290***
(3.57)
1.587**
(2.21)
0.876***
(3.36)
1.071*
(1.86)
0.74
(0.51)
0.013***
(2.93)
0.001
(0.82)
4.491***
(4.18)
357
0.307
0.13
(1.38)
0.145
(0.16)
0.196
(0.57)
0.072
(0.11)
0.007
(0.01)
0.003
(0.56)
0.001
(0.82)
1.823
(1.42)
128
0.061
Panel B. Pre- and post-match pairwise comparisons

Variable

Protability
Assets turnover
Debt-to-assets
Cash-to-assets
Region index
Industry index
Mean
Unmatched
Matched
Unmatched
Matched
Unmatched
Matched
Unmatched
Matched
Unmatched
Matched
Unmatched
Matched
Unmatched
Matched
% reduct.
Treated
Control
%bias
14.068
14.037
0.104
0.105
0.632
0.637
0.226
0.230
0.041
0.041
121.48
117.86
37.708
37.094
13.052
13.625
0.234
0.107
1.136
0.674
0.090
0.204
0.086
0.046
58.812
83.766
28.616
35.953
73.8
29.9
77.9
1.3
92.3
6.9
78.2
14.8
54.4
6.1
45.8
24.9
44.8
5.6
t-Test
Bias
59.5
98.4
92.6
81.1
88.8
45.6
87.5
p N |t|
6.24
1.57
5.46
0.08
5.92
0.56
6.35
0.71
3.46
0.44
3.84
1.3
3.47
0.3
0
0.118
0
0.936
0
0.575
0
0.479
0.001
0.661
0
0.197
0.001
0.766
259
Table 6
Validation test for the regional banking sector dynamics across strategic rms. This table reports difference-in-difference mean estimates of variables across treated and
control groups of rms. Treated rms are interest rate subsidy recipients; controls are non-recipients. The pre-crisis (post-program) growth rates are for the 20062008
(20092011) period. The variables of interest are the density of bank branches per capita and the proportion of regionally incorporated banks in regions where the
strategic rms are located.
Pre-crisis period
Control
Bank branch density

Proportion of regional banks' branches
Treated
Post-crisis period
Diff pre-crisis
Control
D-in-D
Treated
Diff post-crisis
(1)
(2)
(3) = (2) (1)
(4)
(5)
(6) = (5) (4)
(7) = (6) (3)
0.318
(0.068)
0.156
(0.030)
0.298
(0.060)
0.191
(1.46)
0.020
(0.017)
0.035
(0.026)
0.365
(0.073)
0.092
(0.023)
0.339
(0.067)
0.111
(0.023)
0.026
(0.02)
0.019
(0.022)
0.006
(0.007)
0.015
(0.016)
support restriction that drops treatment observations whose propensity score is higher than the maximum or less than the minimum
propensity score of the controls. Our matching produces 64 control rms for 64 treated rms that are on common support.
In order to conduct diagnostics of the propensity score matching validity, we re-estimate the probit model on the matched sample.
The estimation results presented in the second column of Table 5, Panel A show that none of the explanatory variables is signicant
when we restrict the sample to matched rms. Panel B of the Table reveals that average rm characteristics in the matched sample are
not signicantly different between treated and control groups of rms.
In summary, the propensity score diagnostics reported in Table 5 suggests that the matching has removed meaningful
differences in observable rm characteristics during the pre-crisis period that may confound our statistical inference on
estimating the impact of the subsidization program on treated rms. In other words the parallel trends assumption is satised
for the matched sample.
6.2. Validation test for regional banking sector dynamics across strategic rms
The inference in D-in-D regressions may be confounded if variation in banking development is endogenous to unobserved
variation in the investment opportunities of rms. From the perspective of this event study, it is crucial that such simultaneous
development did not occur over the pre- and post-crisis periods on which we focus. The absence of a signicant difference in regional
banking development across treated and control rms over the time period of our study would demonstrate the validity of the
banking development variables that we employ. We use data reported by the Russian Central Bank on the total number of banks
operating in regions and the proportion of regionally incorporated banks. We report these results in Table 6. These estimates are
not signicant suggesting that the level of banking development did not change across treated and control rms over the pre- and
post-crisis periods, thus validating the employment of these variables in our main regressions.
6.3. Difference-in-difference empirical strategy
The primary objective of our study is to determine the impact of the government subsidization program on the corporate
policies of rm-recipients. Because interest rate subsidies increase the free cash ow of rms, we expect that this policy
relaxes the nancial constraints of treated rms and leads to a signicantly different behavior than among the control
group of rms.
We use three sources for identifying variation: 1) rm characteristics during the time before the nancial crisis compared to those
during the post-crisis period after the implementation of the subsidization program; 2) cross section of treated and matched control
rms based on information as to whether they received subsidies or not; and 3) assessment of whether a strategic rm is located in a
city with a high or a low level of banking development. This setup enables us to employ the difference-in-difference (henceforth D-D)
and difference-in-difference-in-difference (henceforth D-D-D) estimation strategies (see Gruber, 1994; Imbens and Wooldridge,
2007; Puri et al., 2011).
We rst start with a classical D-D specication where we examine the response of treated rms to the subsidization program
during the post-program period, where the main coefcient of interest is on an interaction term t TREATi.12
Y i 0 1 TREAT i 2 t 3 t TREAT i i
i indexes rms, t indexes periods, indictor variable TREATi takes a value of 1 if the rm belongs to the treated group that received
government subsidies and 0 if it belongs to the matched control group. The indicator variable t takes a value of 1 if observations
belong to the post-program 20102011 period and 0 if they belong to the pre-program 20062008 period. Because we nonparametrically match control rms on all relevant rm characteristics and satisfy the parallel trends assumption, we do not need
12
We have a two-period panel where the time dimension covers the pre-crisis and post-subsidization periods. This setting greatly reduces concerns about possible
autocorrelation of the dependent variables (see Bertrand et al. (2004)).
260
to include any additional control variables in Eq. (1) (see Lemmon and Roberts, 2010). The dependent variable Yi represents growth
rates of variables studied by Giannetti and Simonov (2013). We calculate them as follows:
Y i ln V 1 V 0 =Total assets0 1 for V Net capex; Total debt; Cash
Y i ln E1 E0 =E0 1 for E Employment:
The growth rates are measured over 20062008 for the pre-crisis period and over 20092011 for the post-subsidization program
period.
Our inference is based on the coefcient 3, which identies the magnitude of the overall impact of the program and captures all
variation in rms' policies specic to the treatments (relative to controls) in the period after the subsidization program (relative to the
period before).
We examine the third source of variation across rms and employ the D-D-D specication:

Y ij 0 1 TREAT i 2 t 3 DEV BANK j 4 t TREAT i 5 t DEV BANK j

6 DEV BANK j TREAT i 7 t TREAT i DEV BANK j ij
where the additional third dimension j indexes cities and the indicator variable DEV_BANKj takes a value of 1 if a strategic rm is located in a city where either the banks' branch density or the proportion of regional bank branches is above their sample medians and 0
otherwise.
In specication (2) the main coefcient of interest is 7 on the third-level interaction terms. It captures all variation in treated
rms' policies relative to control rms in the cities with a high level of banking development (relative to cities with a low level) in
the period after the crisis (relative to the pre-crisis period). The ability to differentiate between different aspects of strategic rms' nancial constraints is a major strength of the three level interactions.
6.4. Estimation results
The estimation results of the D-D and D-D-D models are reported in Table 7. The left panel of the table reports results for the
matched sample, which includes all treated rms that received subsidies and all control rms matched by the PSM algorithm. For a
robustness check, we repeat all our estimations on a sub-sample of nancially distressed rms. Here matching is effectively done by
one characteristic, albeit the key one from the government's perspective.13 The right panel of the table reports the D-D and D-D-D results for the distressed rms.
Column 1 contains estimation results of specication (1) for all PSM matched rms. Next, we partition the PSM matched sample
into subsamples of rms located in cities with high and low densities of bank branches and report D-D estimation for these subsamples in columns 2 and 3. We further partition the matched sample according to the location of rms in cities with a high and
low proportion of regional banks and provide estimates for those respective sub-samples in columns 4 and 5. In the right panel of
Table 7 (columns 610), we follow the same logic for nancially distressed rms.
The rst and second rows of the table report the average growth rate of the nancial variable during the pre-crisis period for
groups of control and treated rms, respectively. These growth rates are captured by the coefcients 0 and 0 + 2 in Eq. (1). For
the post-program period the growth rates for the control and treated groups correspond to the coefcient combinations 0 + 1
and 0 + 1 + 2 + 3 in Eq. (1). We also report differences between the groups during the pre-crisis and post-crisis periods.
At the bottom of the table, we report estimated coefcient 3 for specication (1) in a row labeled D-D. Finally the row
labeled D-D-D corresponds to the estimated coefcient 7 in specication (2). We cluster all standard errors at the regional
level to account for possible within-region correlation between rms.
6.4.1. Capital investment growth
We rst investigate how rms managed their capital spending. As can be seen in the rst two rows of Table 7, the average growth
rate of net capex to initial assets ranged from 0.5 to 2.5% among the studied rms during the pre-crisis period, and the difference between treated and non-treated rms was not signicant. In the post-crisis period, the range of net capex growth rates widened to
0.7%4.8%. Because the subsidization program de jure conditioned receiving rms to invest in capital, one can expect to nd a
high growth of capex spending by treated rms relative to the control rms. However, due to the short-term nature of the program
and the long-term commitment that investment means, we hypothesized that treated rms will not signicantly change their longterm capital spending. Consistent with our fourth hypothesis, we do not nd that treated rms signicantly increased capital spending
relative to control rms as all estimated D-D coefcients from specication (1) are statistically insignicant.14
13
The univariate results reported in the right panel of Table 3 suggest that distressed recipients were not signicantly different from distressed non-recipients along all
key observable characteristics. This means that the parallel trends assumption is satised for this sub-sample.
14
We would like to mention that if only capex expenditures are used instead of net capex spending in Eq. (1), the results for the D-D estimation are signicant and
point at a higher capex spending by treated rms relative to control rms. These results are available from the authors upon request.
261
Table 7
Difference-in difference tests for the interest rate subsidy recipients before and after government subsidization program (Cont.).
The table reports difference-in-difference mean estimates of variables across treated and control groups of rms. Treated rms are interest rate subsidy recipients; controls are non-recipients. The pre-crisis (post-program) growth rates are for the 2006-2008 (2009-2011) period. The left panel reports results for all subsidy recipients
and matched control rms. The right panel reports results for nancially distressed rms. Each sample is partitioned by two characteristics of the banking sector in the
cities where rms are located: bank branch density per capita; and the proportion of regional bank branches.
Financially distressed rms
Subsidy recipients and matched non-recipients

Full
matched
sample
Low
branch
density
High
branch
density
Low
region.
bank
High
region.
bank
Full n.
distressed
sample
Low
branch
density
High
branch
density
Low
region.
bank
High
region.
bank
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
(9)
(10)
0.015 (0.95)
0.024
(1.16)
0.019
(0.37)
-0.005
(0.22)
0.040
(0.43)
-0.007
(0.50)
-0.047
(1.07)
-0.041
(1.46)
0.070 (1.65)
0.012
(0.60)
0.009
(0.26)
-0.003
(0.14)
0.013
(0.10)
0.038
(1.18)
0.025
(1.09)
0.028
(0.78)
0.005
0.021
(0.19)
(1.11)
0.018
0.008
(1.10)
(1.08)
0.012
-0.013
(0.45)
(0.56)
0.038
0.008
(1.18)
(1.12)
0.003
0.042
(0.37)
(1.62)
-0.035
0.034
(1.50)
(1.56)
-0.048
0.048
(1.07)
(1.27)
0.095* (1.84)
Net capex growth over initial assets

Pre-crisis Control
0.010
period
rms
(1.12)
Treated
0.013
rms
(3.63)
Difference
0.002
pre-crisis
(0.23)
PostControl
0.027
crisis
rms
(2.03)
period
Treated
0.040
rms
(2.70)
Difference
0.012
Post-crisis
(0.62)
D-in-D
0.010
(0.48)
D-in-D-in-D
0.021
0.006
(1.84)
(0.53)
0.026
0.008
(2.54)
(2.63)
0.006
0.002
(0.014)
(0.13)
0.006
0.035
(0.77)
(2.89)
0.022
0.046
(0.64)
(2.47)
0.016
0.010
(0.44)
(0.42)
0.010
0.009
(0.33)
(0.32)
-0.001 (0.03)
0.012
0.010
(0.64)
(0.91)
0.026
0.008
(2.35)
(3.94)
0.014
-0.002
(0.74)
(0.19)
0.024
0.028
(0.55)
(2.21)
0.019
0.048
(1.32)
(2.05)
-0.005
0.019
(0.72)
(0.97)
-0.020
0.022
(0.49)
(0.79)
0.041 (0.87)
0.032
(3.83)
0.053
(0.60)
0.021
(0.55)
0.053
(3.19)
0.032
(1.60)
-0.021
(1.13)
-0.042
(1.35)
-0.011
0.045
(0.37)
(4.33)
0.030
0.065
(1.69)
(0.57)
0.040
0.020
(1.35)
(0.46)
0.021
0.079
(0.85)
(1.22)
0.021
0.047
(0.63)
(1.95)
-0.001
-0.032
(0.57)
(1.1)
-0.040
-0.052
(0.61)
(1.23)
-0.011 (0.14)
0.029 (0.75)
D-in-D-in-D
0.027
(0.71)
0.058
(1.47)
0.031
(0.70)
0.096
(1.09)
0.063
(0.74)
-0.033
(0.55)
0.063
(0.60)
0.021 (0.19)
Total debt growth over initial assets

Pre-crisis Control
0.066
period
rms
(3.78)
Treated
0.133
rms
(3.33)
Difference
0.068**
pre-crisis
(2.27)
PostControl
0.116
crisis
rms
(1.69)
Treated
0.102
period
rms
(2.07)
Difference
-0.013
Post-crisis
(0.38)
D-in-D
-0.081*
(1.80)
D-in-D-in-D
0.022
(0.64)
0.106
(2.06)
0.084
(1.45)
0.124
(1.77)
0.097
(1.59)
-0.028
(1.25)
-0.111
(1.33)
0.043 (0.41)
0.081
(4.46)
0.141
(2.57)
0.060
(1.68)
0.112
(0.94)
0.104
(1.37)
-0.009
(1.53)
-0.069
(1.22)
0.012
0.083
(0.36)
(4.35)
0.092
0.149
(2.10)
(3.71)
0.080
0.066*
(1.49)
(1.91)
0.124
0.113
(1.67)
(0.95)
0.151
0.083
(0.79)
(2.04)
0.027
-0.030
(0.52)
(0.458)
-0.053
-0.095**
(0.48)
(2.31)
-0.042 (0.37)
0.055 (1.97)
Cash growth over initial assets

Pre-crisis Control
0.030
period
rms
(3.23)
Treated
0.035
rms
(0.64)
Difference
0.005
pre-crisis
(0.47)
Control
0.090
rms
(3.57)
0.008
(0.43)
0.007
(0.45)
-0.001
(0.04)
0.069
(2.28)
0.038
(5.37)
0.043
(0.63)
0.006
(0.49)
0.097
(2.56)
0.015
(0.90)
0.009
(0.84)
-0.006
(0.27)
0.091
(2.57)
Employment growth
Pre-crisis Control
period
rms
Treated
rms
Difference
pre-crisis
PostControl
crisis
rms
period
Treated
rms
Difference
Post-crisis
D-in-D
0.031
(2.66)
0.054
(1.84)
0.024
(0.74)
0.065
(3.55)
0.040
(1.92)
-0.025
(1.31)
-0.048
(1.47)
0.035
(4.74)
0.045
(1.03)
0.010
(0.81)
0.089
(2.62)
0.011 (0.41)
-0.004 (0.21)
0.019 (0.36)
0.028 (0.65)
0.008 (0.53)
0.012 (0.41)
0.033 (0.10)
0.004 (0.05)
0.062 (0.73)
-0.009 (1.49)
-0.071 (1.07)
-0.074 (1.10)
0.159 (2.76)
0.104** (2.33)
0.097 (0.67)
0.107 (1.42)
0.010 (0.87)
-0.094 (1.01)
0.025 (3.94)
0.042 (0.85)
0.017 (0.77)
0.079 (2.22)
-0.002
0.039
(0.02)
(0.42)
0.046
0.029
(1.33)
(0.15)
0.048
-0.011
(0.40)
(0.14)
-0.012
0.085
(0.09)
(0.98)
0.085
-0.042
(0.31)
(2.87)
0.096
-0.127*
(0.28)
(1.80)
0.048
-0.116
(0.24)
(1.30)
-0.164 (0.75)
-0.037
0.070
(0.77)
(1.19)
0.001
0.057
(1.43)
(0.09)
0.037
-0.013
(0.71)
(0.12)
0.115
0.030
(1.35)
(1.09)
-0.052
0.022
(3.39)
(0.10)
-0.166
-0.007
(1.59)
(0.06)
-0.204**
0.006
(2.19)
(0.07)
0.210 (1.61)
0.029
0.063
(0.96)
(1.81)
0.158
0.159
(1.31)
(2.41)
0.129
0.096*
(1.22)
(1.94)
-0.002
0.128
(2.90)
(0.75)
0.137
0.098 (1.8)
(0.19)
0.139
-0.029
(0.19)
(0.97)
0.010
-0.126
(0.06)
(1.15)
-0.136 (0.79)
0.026
0.073
(1.03)
(1.82)
0.116
0.183
(1.36)
(2.46)
0.090
0.111*
(1.17)
(1.92)
0.032
0.136
(0.15)
(0.64)
0.130
0.094
(0.21)
(1.74)
0.098
-0.043
(0.16)
(0.97)
0.008
-0.154
(0.06)
(1.25)
-0.161 (0.89)
0.021
(1.15)
0.008
(3.87)
-0.013
(0.70)
0.069
(0.72)
0.017
(1.52)
0.001
(3.98)
-0.016
(1.30)
0.037
(0.48)
0.026
(3.64)
0.052
(1.04)
0.026
(0.94)
0.082
(1.87)
0.029
(3.64)
0.065
(1.22)
0.036
(1.15)
0.105
(2.69)
262
Table 7 (continued)
Financially distressed rms
Subsidy recipients and matched non-recipients
Postcrisis
period
Treated
rms
Difference
Post-crisis
D-in-D
D-in-D-in-D
N. obs.
Full
matched
sample
Low
branch
density
High
branch
density
Low
region.
bank
High
region.
bank
Full n.
distressed
sample
Low
branch
density
High
branch
density
Low
region.
bank
High
region.
bank
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
(9)
(10)
0.033
(4.82)
-0.057***
(3.34)
-0.062**
(2.64)
0.088
0.016
(0.59)
(8.95)
0.018
-0.081***
(0.43)
(3.90)
0.019
-0.087***
(0.45)
(3.08)
-0.106** (2.23)
64
192
-0.003 (8.07)
0.024
-0.011
(2.13)
(8.72)
-0.045
-0.093**
(0.47)
(3.50)
-0.033
-0.119**
(0.61)
(2.62)
-0.087 (1.25)
20
66
256
0.071
0.018
(0.31)
(8.11)
-0.021
-0.071***
(0.29)
(4.20)
-0.015
-0.081***
(0.25)
(3.15)
-0.067 (1.07)
68
188
-0.082***
(3.55)
-0.099**
(2.64)
86
0.017
-0.014
(0.54)
(9.14)
-0.021
-0.119***
(0.12)
(4.54)
-0.005
-0.155***
(0.14)
(3.11)
-0.150** (2.55)
32
54
Notes: *, **, *** Denote signicance at 10%; 5%; and 1% respectively and are reported for difference estimates only. All standard errors are clustered at the region level.
Next we turn to the estimate of the triple-interaction term from the specication (2) and test if variation with regard to banking development in cities where rms are located affects their capital spending. We nd a marginal signicance (at 10%) of D-DD coefcients for the sub-sample of nancially distressed rms. The economic magnitude of the obtained coefcients is rather
high: 7% for variability by bank branch density and 9.5% for variability by the proportion of regional banks. These results show
that distressed rms located in cities with low banking development exhibited a lower investment rate than control rms in
similar locations during the post-crisis period. On the contrary, treated rms located in cities with high banking development
invested more than control rms in similar locations during this period.
In sum, our last nding suggests that the government subsidization program had a positive impact on investments by the
recipients provided that they were located in areas with a high level of banking development, which served as a proxy of the nancial
constraints of rms.
6.4.2. Employment growth
Another socially important variable that we study is employment growth. When we examined the rationale of the government's
subsidization program, we found that it had an implicit agenda of supporting employment in economically vulnerable parts of the
country.
The estimation of Eq. (1) for employment growth produces negative D-D coefcients for most samples we examine. This
means that treated rms had slower average growth of employment relative to control rms in the post-crisis period. All coefcients except for one are statistically insignicant, which suggests that the subsidization program did not have any pronounced
impact on rms' employment policies. The statistical signicance of the estimated D-D-D coefcients for employment growth is
also low.
6.4.3. Debt management
The results for this variable reveal that while treated and control groups of rms started at the same average level of debt-to-assets
ratio produced by the PSM algorithm, treated rm exhibited a higher growth of debt during the pre-crisis period. We can see that the
average growth rates of debt equalized across two groups during the post-crisis period. The D-D coefcients are negative and marginally signicant; however, given the fact that the results are driven by the signicant difference during the pre-crisis period, one should
interpret them with caution.
Moving to specication (2) one could expect that given the differences in banking development in cities where rms are located, there is a differential supply of credit across examined groups of rms that translates into differences in total debt growth
of rms. We do not nd this to be the case as all estimated D-D-D coefcients are insignicant. This result could be tentatively
interpreted to imply that the subsidization program helped to sustain borrowing by nancially constrained rms across all
locations.
6.4.4. Cash management
Our Hypothesis 5 posits that rms manage their cash in a way so as to offset possible credit frictions. The rst two rows of the table
report the average growth rates of cash holdings during the pre-crisis period. One can see that growth rates of cash holdings are similar across treated and control groups for all studied samples of rms.
As visible from column 1 of the table, the cash growth rate of treated rms remained at the pre-crisis level, while it signicantly
accelerated among control group rms in the period after the Lehman Brothers collapse. This nding is consistent with the theory
of precautionary cash hoarding by rms (Han and Qiu (2007)). The interest payment subsidization program freed up the available
free cash-ow of treated rms and enabled them to maintain cash holdings at the pre-crisis level. At the same time, control group
rms that were hit by the external credit supply shock without receiving any government assistance started accumulating cash, as
263
the theory predicts. The D-D coefcient estimate shows that control rms increased their pace of cash accumulation by 6% relative to
treated rms in the post-program period. This nding conrms our fth hypothesis.
Interesting estimation results emerge for sub-samples partitioned by the level of banking development in the cities where strategic rms are located. Columns 2 and 4 of the table reports the estimates for the subsample of rms that are classied as being located
in areas with low banking development by our two banking development proxies. One can see that both groups of rms (treated and
control) exhibited equally high growth rate of cash holding in the post-program period. This means that in cities with low banking
development rms felt nancially constrained regardless of whether they received government subsidies. When we examine columns 3 and 5 that show the estimation results for rms located in cities with high banking development, we nd the same differential
pace of cash accumulation between treated and non-treated rms that we found for the full sample (in column 1) in the post-crisis
period.
Testing the specication (2), we nd that both D-D-D coefcient estimates are negative but that only the one that captures the
differences in bank branch density is statistically signicant. Thus, there was a signicant 10.6% drop in the relative growth of cash
holdings of treated rms after the crisis and the completion of the subsidization program in cities with a high bank branch density
relative to control rms in cities with a low branch density.15
It is reassuring that all the results for nancially distressed rms reported in the right panel of the table are qualitatively similar to
those for the PSM matched sample that we described above. Thus, we cannot reject Hypothesis 6.
7. Conclusion
What nancial and socioeconomic factors determine the choice of rms to which government subsidies are allocated? Do government subsidies have a pronounced impact on the corporate policies of rms that benet from such assistance? In this study we have
addressed these questions by taking advantage of a unique dataset on the debt subsidization program that the Russian government
implemented to support strategic rms during the 2009 crisis.
First, we looked at the rationale of the government subsidization program. Our ndings suggest that the worse a rm performed
during the pre-crisis period, the more likely it was to be selected for assistance. We also nd that the likelihood that a rm received
subsidies is related to its size and location, with most aid recipients being large rms based in small cities with few employment alternatives. Taken together, these results suggest that the implementation of the debt subsidization program was largely motivated by
political considerations and aimed at preventing bankruptcies of strategic rms and potential social unrest that could result from such
bankruptcies in economically vulnerable parts of the country. These ndings shed light on the implicit agenda of government programs designed to extend a helping hand to the private sector.
Second, we investigated the post-crisis performance of rms that received government assistance compared to that of rms that
did not benet from this program. Employing an empirical difference-in-difference strategy, we show that the program did not have a
signicant impact on capital investments of subsidy recipients, contrary to what could be expected given program priorities. Nevertheless, we can still tentatively conclude that the outcome of the program was positive since there were no bankruptcies among the
recipients. We also nd that the matched group of non-recipients on average exhibited a higher degree of cash hoarding behavior than
subsidy-recipients in the post-program period, which suggests that the program eased external nancial constraints of recipientrms. Consistent with the theory on rms' precautionary cash savings, we further nd that non-recipients based in cities with a
low banking development accumulated more cash holdings than subsidy-recipients based in cities with a high level of banking
development.
As a whole, our ndings provide a contribution to the theoretical and empirical literature on investment and cash management policies of rms under nancial constraints. Our results are consistent with theoretical predictions and indicate that additional short-term cash ows have a marginal impact on the long-term investment decision of rms and that the cash
holding behavior of rms depends on the level of nancial constraints they face (the less constraints they face, the less cash
they accumulate).
Acknowledgment
Vladimir Sokolov acknowledges the nancial support of the NRU Higher School of Economics Center for Advanced Studies
(CAS) individual research grant. We thank Yuriy Gorodnichenko, Chris Julliard, William Megginson, Thomas Noe, Udara
Peiris, Andrei Simonov, Laura Solanko, three anonymous referees, and participants at the 2013 ACES Meeting in San Diego,
the 2013 iCare conference in Perm and research seminars at the Bank of Finland and ICEF for valuable comments and
suggestions.
15
In the unreported results, we apply D-D and D-D-D to the average cashassets ratio and nd similar differential cash hording behavior across treated and control
rms.
264
Appendix A
Table A1
Variable denitions and data sources.
Dependent variables
Subsidy recipient
Subsidy-to-assets
Net capex to initial assets
growth
Employment growth
Total debt to initial assets
growth
Cash to initial assets growth
Independent variables
Total revenue
Tangibility
Debt-to-assets
Protability
Cash holdings
Assets turnover
State ownership share
City population
Distance to capital
Average year temperature
HHI employment
Ruling party concentration
High ways length per area
Bank branch density
Proportion of regional banks
Denition
Source
Unit
The variable takes value one if a strategic rm was a subsidy recipient in 2009, 0 otherwise
The ruble value of a subsidy over average 20062007 total assets
Change in the net capital expenditures between periods 0 and 1 scaled by period 0
total assets
Change in rm's employment between periods 0 and 1 scaled by period 0 employment
Change in total debt between periods 0 and 1 scaled by period 0 total assets
MED of RF
MED of RF
Ruslana, BvD
(0,1)
%
%
Ruslana, BvD
Ruslana, BvD
%
%
Change in cash holdings between periods 0 and 1 scaled by period 0 total assets
Ruslana, BvD
Total sales of a rm
Tangible xed assets divided by total assets
Total debt to total assets
Operating income before interest and taxes scaled by total assets
Cash and cash equivalents divided by total assets
Total revenue to total assets
Share of a strategic rm owned by the federal government
Population of a city where a strategic rm is located
Distance between Moscow and the city where a strategic rm is located
Average annual temperature in a city where a strategic rm is located
Herndahl index (HHI) of employment concentration for each city
Number of deputies from a region representing the United Russia/the region's population
Total length of highways in a region where a strategic rm is located per region's area
Total number of bank branches per capita in a city where a strategic rm is located
Number of branches of regional banks in relation to the overall number of banks in
the city
Ruslana, BvD
Ruslana, BvD
Ruslana, BvD
Ruslana, BvD
Ruslana, BvD
Ruslana, BvD
Ruslana, BvD
Rosstat
Rosstat
Rosstat
Ruslana, BvD
Wikipedia
Rosstat
BEPS II, EBRD; CBR
BEPS II, EBRD; CBR
Mln. RUB
%
%
%
%
%
%
Thousands
km
C
Units
Numb. per mln.
km per km2
Units
Share
Notes: MED of RF and Rosstat are the Ministry of Economic Development of the Russian Federation and the Federal State Statistics Service of the Russian Federation,
respectively. Ruslana, BvD is the database provided by the Bureau van Dijk; BEPS II, EBRD Banking Environment and Performance Survey II; CBR is the Central Bank
of Russia.
References
Almeida, H., Campello, M., Weisbach, M., 2004. The cash flow sensitivity of cash. J. Financ. 59, 17771804.
Almeida, H., Campello, M., Weisbach, M., 2011. Corporate financial and investment policies when future financing is not frictionless. J. Corp. Financ. 17, 675693.
Almeida, H., Campello, M., Laranjeira, B., Weisbenner, S., 2012. Corporate debt maturity and the real effects of the 2007 credit crisis. Crit. Financ. Rev. 1, 358.
Andrade, G.,Kaplan, S., 1998. How costly is financial (not economic) distress? Evidence from highly leveraged transactions that became distressed. J. Financ. 53, 14431493.
Asquith, P., Gertner, R., Sharfstein, D., 1994. Anatomy of financial distress: an examination of junk-bond issues. Q. J. Econ. 109, 625658.
Benfratello, L., Sciantarelli, F., Sembenelli, A., 2008. Banks and innovation: microeconometric evidence on Italian firms. J. Financ. Econ. 90, 197217.
Bertrand, M., Duflo, E., Mullainathan, S., 2004. How much should we trust difference-in-difference estimates? Q. J. Econ. 249275.
Bharath, S., Dahiya, S., Saunders, A., Srinivasan, A., 2007. So what do i get? The bank's view of lending relationship. J. Financ. Econ. 85, 368419.
Bircan, C., De Haas, R., 2014. Bank lending and firm innovation: evidence from Russia, mimeo.
Boubakri, N., Cosset, J.-C., Guedhami, O., 2009. From state to private ownership: issues from strategic industries. J. Bank. Financ. 33, 367379.
Brown, M., De Haas, R., Sokolov, V., 2013. Regional inflation and financial dollarization. EBRD Working Paper 163.
Bruno, R.L., Bytchkova, M., Estrin, S., 2013. Institutional determinants of new firm entry in Russia: a cross-regional analysis. Rev. Econ. Stat. 95, 17401749.
Campello, M., Graham, J., Harvey, C., 2010. The real effects of financial constraints: evidence from a financial crisis. J. Financ. Econ. 97, 470487.
Caselli, F., Michaels, G., 2013. Do oil windfalls improve living standards? Evidence from Brazil. Am. Econ. J. Appl. Econ. 5, 208238.
Commander, S., Nikoloski, Z., Plekhanov, A., 2011. Employment concentration and resource allocation: one-company towns in Russia. EBRD Working Paper 130.
Cull, R., Xu, L., 2005. Institutions, ownership, and finance: the determinants of profit reinvestment among Chinese firms. J. Financ. Econ. 77, 117146.
Dasgupta, S., Noe, T., Wang, Z., 2011. Where did all the dollars go? The effect of cash flows on capital and asset structure. J. Financ. Quant. Anal. 46, 12591294.
Degryse, H., Ongena, S., 2005. Distance, lending relationships, and competition. J. Financ. 60, 231266.
Denis, D., Sibilkov, V., 2010. Financial constraints, investments, and the value of cash holdings. Rev. Financ. Stud. 23, 247269.
Di Patti, E.B., Dell'Ariccia, G., 2004. Bank competition and firm creation. J. Money, Credit, Bank. 36, 225251.
Duchin, R., Sosyura, D., 2011. Safer ratios, riskier portfolios: banks' response to government aid. Mimeo.
Duchin, R., Sosyura, D., 2012. The politics of government investment. J. Financ. Econ. 106, 2448.
Duchin, R., Ozbas, O., Sensoy, B., 2010. Costly external finance, corporate investment, and the subprime mortgage crisis. J. Financ. Econ. 97, 418435.
Faccio, M., Masulis, R., McConnell, J., 2006. Political connections and corporate bailouts. J. Financ. 61, 25972635.
Faulkender, M., Wang, R., 2006. Corporate Financial Policy and the Value of Cash. Journal of Finance 61, 19571990.
Frye, T., Shleifer, A., 1997. The invisible hand and the grabbing hand. Am. Econ. Rev. 87, 354358.
Galpin, R., 2009. Unrest Threat as Crisis Hits Russia. BBC News (http://news.bbc.co.uk/2/hi/europe/7973084.stm).
Giannetti, M., Simonov, A., 2013. On the real effects of bank bailouts: micro evidence from Japan. Am. Econ. J. Macroecon. 5, 135167.
Gruber, J., 1994. The incidence of mandated maternity benefits. Am. Econ. Rev. 84, 622641.
Han, S., Qiu, J., 2007. Corporate precautionary cash holdings. J. Corp. Financ. 13, 4357.
Imbens, G., Wooldridge, J., 2007. Difference-in-difference estimation, mimeo.
Jayarathe, J., Strahan, P., 1996. The financegrowth nexus: evidence from bank branch deregulation. Q. J. Econ. 111, 639670.
265
Lemmon, M., Roberts, M., 2010. The response of corporate financing and investment to changes in the supply of credit. J. Financ. Quant. Anal. 45, 555587.
Lerner, J., 1999. The government as venture capitalist: the long-run impact of the SIBR program. J. Bus. 72, 285318.
Megginson, W., Netter, J., 2001. From state to market: a survey of empirical studies on privatization. J. Econ. Lit. 39, 321389.
Megginson, W., Nash, R., Netter, J., Poulsen, A., 2004. The choice of private versus public capital markets: evidence from privatizations. J. Financ. 59, 28352870.
Philippon, T., Schnabl, P., 2013. Efficient recapitalization. J. Financ. 68, 142.
Puri, M.,Rocholl, J.,Steffen, S., 2011. Global retail lending in the aftermath of the US financial Crisis: distinguishing between supply and demand effects. J. Financ. Econ.
100, 556578.
Roberts, M.R., Whited, T., 2012. Endogeneity in empirical corporate finance. In: Constantinides, G., Harris, M., Stulz, R. (Eds.), Handbook of the Economics of Finance,
Vol. 2. Elsevier Science, North Holland.
Shleifer, A., Vishny, R., 1994. Politicians and firms. Q. J. Econ. 109, 9951025.
Sokolov V., 2012. Tracing the impact of central bank liquidity infusions on financially constrained banks: evidence from a natural experiment, mimeo.
Zhao, Z., 2004. Using matching to estimate treatment effects: data requirements, matching metrics, and Monte Carlo evidence. Rev. Econ. Stat. 86, 91107.

Journal of Empirical Finance: Yulia Davydova, Vladimir Sokolov

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Journal of Empirical Finance: Yulia Davydova, Vladimir Sokolov

Uploaded by

Copyright:

Available Formats

Journal of Empirical Finance 29 (2014) 247265

Contents lists available at ScienceDirect

Journal of Empirical Finance

The real effects of nancial constraints: Evidence from a debt

CitiGroup Global Markets Prime Finance Risk, London, UK

Y. Davydova, V. Sokolov / Journal of Empirical Finance 29 (2014) 247265

Y. Davydova, V. Sokolov / Journal of Empirical Finance 29 (2014) 247265

Y. Davydova, V. Sokolov / Journal of Empirical Finance 29 (2014) 247265

2.2. The impact of the program on rms' corporate policies

3. Background of the debt subsidization program

Y. Davydova, V. Sokolov / Journal of Empirical Finance 29 (2014) 247265

Fig. 1. Dynamics of the seasonally adjusted manufacturing index.

3.2. Subsidization of the strategic rms program

Y. Davydova, V. Sokolov / Journal of Empirical Finance 29 (2014) 247265

Naonal Reserve Fund of the Russian

Fig. 2. Dynamics of the National Reserve Fund.

4.2. Description of variables

4.2.1. Financial characteristics of rms

4.2.2. Geographic, socio-economic and political characteristics of cities

Y. Davydova, V. Sokolov / Journal of Empirical Finance 29 (2014) 247265

ISIC Rev. 4 codes

Total number of Number of rms receiving Interest payment subsidies/

4.2.3. Banking development characteristics

Y. Davydova, V. Sokolov / Journal of Empirical Finance 29 (2014) 247265

4.3. Summary statistics and univariate analysis

Y. Davydova, V. Sokolov / Journal of Empirical Finance 29 (2014) 247265

5. The determinants of the government subsidization program

Probit (marginal effects)

Firm size (log revenue)

Y. Davydova, V. Sokolov / Journal of Empirical Finance 29 (2014) 247265

5.1. Probit results

Y. Davydova, V. Sokolov / Journal of Empirical Finance 29 (2014) 247265

Firm size (log revenue)

Panel B. Pre- and post-match pairwise comparisons

Firm size (log revenue)

Y. Davydova, V. Sokolov / Journal of Empirical Finance 29 (2014) 247265

Bank branch density

(3) = (2) (1)

(6) = (5) (4)

(7) = (6) (3)

Y. Davydova, V. Sokolov / Journal of Empirical Finance 29 (2014) 247265

Y. Davydova, V. Sokolov / Journal of Empirical Finance 29 (2014) 247265

Subsidy recipients and matched non-recipients

Net capex growth over initial assets

Total debt growth over initial assets

Cash growth over initial assets

Y. Davydova, V. Sokolov / Journal of Empirical Finance 29 (2014) 247265

Subsidy recipients and matched non-recipients

Y. Davydova, V. Sokolov / Journal of Empirical Finance 29 (2014) 247265

Y. Davydova, V. Sokolov / Journal of Empirical Finance 29 (2014) 247265

Y. Davydova, V. Sokolov / Journal of Empirical Finance 29 (2014) 247265

You might also like