Bank Lending and Performance of Micro, Small and Medium Sized Enterprises (MSMEs):
Evidence from Bulgaria, Georgia, Russia and Ukraine
Karin Jõeveer*
Francesca Pissarides**
Jan Svejnar***
March 2006
* Keele University
** EBRD
*** University of Michigan and CERGE-EI
The paper was written with a financial and institutional support of the Japan Europe Development Fund and EBRD. We would like to thank members of the Office of the Chief Economist at EBRD for useful comments. The usual disclaimer applies.
2
Bank Lending and Performance of Micro, Small and Medium Sized Enterprises (MSMEs):
Evidence from Bulgaria, Georgia, Russia and Ukraine 1. Introduction
In view of the important contribution which entrepreneurs and micro, small
and medium-sized enterprises (MSMEs) can make to economic growth, innovation
and employment creation, both researchers and policy makers emphasise the need to
obtain a better understanding of the factors that influence the rise and performance of
these firms. Academic research has identified particular constraints on the availability
of finance for MSMEs such as informational asymmetries between borrowers and
lenders, lack of credit history on the part of many MSMEs, scarcity of appropriate
credit skills in banks, and economies of scale in lending. To overcome these
impediments, many governments, international financial institutions and non-
government organizations (NGOs) have established programmes that target the
delivery of medium- to long-term credit to MSMEs through financial intermediaries.
The European Bank for Reconstruction and Development (EBRD), being the largest
development finance lender in the transition economies and one of the largest in the
world, has for instance been implementing micro and SME lending programmes that
aim to build credit skills for MSME lending in existing participating banks (PBs) and
newly established specialised banks known as microfinance institutions (MFIs).
EBRD’s lending also aims to develop PBs’ credit procedures that reduce lending costs
and to help borrowers build a credit history and lower banks’ perceptions of risk
associated with this type of lending. Interestingly, while the objectives of MSME
lending programmes are widely accepted a being important, little evidence is
available on the impact and longer-term financial sustainability of these programmes
3
(see e.g., Hulme and Moseley, 1996, Morduch, 1999, and Brown, Earle and Lup,
2002).
The purpose of this paper is to assess the impact on MSME performance of the
provision of short-, medium- and long-term credit by banks to these firms. During the
period covered by our study, standard bank credit in the transition economies under
consideration was of a short-term nature, focusing on providing the firms with
working capital. There was hence an important gap in the financial market and
anecdotal evidence suggests that firms sometimes tried to use a series of short term
loans to finance longer term capital investments. In view of this challenge, EBRD
started providing medium- to long-term credit to enable the firms to finance capital
investment and other forms of longer term restructuring. In order to carry out our
analysis, in 2005 we administered a survey to a sample of firms that had received a
loan from the EBRD MSME lending programmes in 2002 and to a sample of similar
firms that had never received an EBRD program loan. The latter sample represents
our control group. In both groups, some firms had received short-term loans from
non-EBRD sources prior to 2002 and some had not. In the survey, we obtain this
information as well as data on EBRD and non-EBRD loans that the firms obtained
between 2002 and 2004, as well as performance indicators for all firms between 2002
and 2004. A more detailed discussion of the sample is provided below.
There are two key questions that we address in this research. First, did
MSMEs that had received short-term (non-EBRD) versus medium- to long-term
(EBRD) loans prior to 2002 subsequently attain greater recourse to bank finance than
firms that had not received bank credit prior to 2002? Second, what has been the
effect of short- versus longer-term credit on MSME performance? In order to provide
a relatively comprehensive understanding, we use several indicators of firm
4
performance: survival, investment, revenues, profit, labour cost, employment, and
market share.
The paper is organised as follows. Section 2 describes the EBRD financing
projects for MSMEs, while Section 3 outlines the hypothesized effects of these
programs. Section 4 discusses the main features of the survey and basic statistics,
while Section 5 presents the analytical framework that we use. The empirical results
are discussed in Section 6 and the conclusions are drawn in Section 7.
2. EBRD projects targeting provision of finance to MSMEs
One of EBRD’s operational priorities is to support MSMEs in its region of
operation with medium- to long-term loans because it regards the lack of longer
maturity loans (i.e., those for fixed as opposed to only working capital investment) in
as resulting in a suboptimal scale of MSME activity. There are two important market
failures that EBRD addresses in this respect. One concerns the inadequate incentive
structure to allow for capacity building within the MSME lending institutions. The
second failure results in an underdeveloped culture of credit in the MSME segment of
the banking market on both sides of the market.
The EBRD pursues its objective through the provision of credit via financial
intermediaries and through the channelling of technical assistance funds to these
financial intermediaries or directly to the MSMEs. In particular, based on the belief
that the provision of lending to this group of enterprises is still inadequate, the EBRD
provides the aforementioned (indirect) MSME lending as an additional instrument
beyond its support to foreign banks’ entry and bank privatisation to accelerate the
development of this specific type of finance. This intervention by the EBRD is
designed to bring benefits to both the MSMEs and the financial institutions that
5
engage in MSME financing. The support is designed with a view to make these
activities financially sustainable in the long run. Microfinance institutions of the type
supported by EBRD operate on a commercial basis by providing loan finance to
clients that profit-maximising financial institutions would not yet serve (or not on an
adequate scale) based on transactions costs and risk/return considerations.
The paper evaluates the overall impact of three types of EBRD financing
operations aimed at supporting MSMEs: Those with (1) de novo dedicated micro-
finance banks, (2) existing banks participating in broad micro-lending programmes
and (3) existing banks participating in the EU/EBRD-SME Facility.1
Microfinance banks are set up by both private and public shareholders to
provide finance on a purely commercial basis to micro and small businesses. Most –
but not all – shareholders of the micro-finance banks are multilateral institutions (such
as the EBRD), bilateral donors, not-for profit private charities and NGOs. These
banks benefit from technical assistance to finance initial set up costs and later branch
expansion. Among all EBRD programmes targeting the provision of finance to
MSMEs, microfinance banks are associated with the greatest scale of activity.2 The
rationale for establishing the initially costly microfinance banks, as opposed to
working with existing local partner banks, is to create a reliable, permanent delivery
mechanism for MSME finance. The microfinance banks can also play an important
part in financial sector development by demonstrating the commercial viability of
MSME lending to other market participants. The EBRD microfinance banks were set
up under two different sets of circumstances: a) post-war reconstruction situations
1 Due to a limited sample size, in our estimations we treat these three programs as a single initiative (we check, however, whether one particular bank -- Hebros in Bulgaria -- yields different results. EBRD also uses SME credit lines within other programmes, stand-alone SME credit lines, credit lines to leasing companies within the EU/EBRD SME Facility, dedicated SME equity funds and, in special cases, direct equity investments in SMEs. 2 EBRD (2004), “Transition Impact and Subsidies in the ERBD’s Micro, Small and Medium-Sized Enterprise Financing Operations”.
6
and/or dysfunctional crisis-ridden financial sectors; b) lack (or under-performance) of
viable local banks.
When and where it is possible, carefully selected local banks are provided by
EBRD with medium- to long-term finance which is then on-lent to small and
medium-sized enterprises. The provision of credit lines is accompanied by technical
assistance which finances capacity building.3 This is the case both for existing banks
participating in broad micro-lending programmes and for existing banks participating
in the EU/EBRD-SME Facility.4 In addition to the technical assistance component,
banks participating in the EU/EBRD-SME Facility receive a performance fee
associated with the provided credit line. This fee is a conditional subsidy – a discount
on the interest rate charged by the EBRD to participating bank at the end of each
interest payment period, on outstanding amounts drawn down from the Facility’s
credit lines.5
The subsidies provided to the financial intermediaries participating in EBRD
MSME programmes are not transferred to the MSMEs. The effective rates charged by
EBRD sponsored microfinance banks and banks participating in EBRD programmes
are on average in line with the rates charged by their competitors.
In terms of collateral requirements, however, there are significant differences
between what microfinance banks and banks participating in EBRD micro-lending
programmes accept from their clients as collateral as compared to what other banks
(including banks participating in the EU/EBRD-SME Facility) require. Both 3 The exact elements of capacity building typically vary from case to case ranging from investments in skills, branch expansion and information technology 4 EBRD also uses SME credit lines within other programmes, stand-alone credit lines and credit lines to leasing companies within the EU/EBRD SME Facility. 5 The discount on the interest rate is supposed to be granted only on the condition that the amounts of the credit line drawn down by the financial intermediaries satisfy the following conditions: (1) the funds need to be used to provide finance to MSMEs with ceilings both on the size of the loan and the size of the enterprise eligible for loans; (2) the final beneficiaries must be new clients of the participating financial institutions; and (3) the quality of the loans/leases made by the participating financial institutions must be of at least a certain standard (measured by arrears).
7
microfinance banks and banks participating in micro-lending programmes managed
by EBRD use very flexible definitions of collateral that allows MSMEs, which would
not have otherwise been able to access bank credit, to benefit from bank lending.
In terms of maturity, banks participating in the EU/EBRD SME Facility tend
to offer the longest terms for their loans. These banks offered loans with an average
27 months maturity versus an average of 18 months offered by microfinance banks
and 12 months offered by banks participating in micro-lending programmes. Banks
which were not participating in EBRD MSME programmes usually extend loans of
shorter maturity than banks participating in EBRD programmes to the same size
category of clients.
The beneficiaries of the first two types of EBRD programmes are private
entrepreneurs and enterprises, ranging from self employed one person businesses to
companies with up to 100 employees. In order to create substantial access to finance
for MSEs, all sectors of the economy in as many regions as possible are targeted,
independent of the size of the loan required. Loans start as low as USD 20 (e.g. for an
open bakery on a Central Asian market to buy flour) up to about USD 200,000 (e.g.
for the purchase of upholstery equipment for a furniture producer in Ukraine). The
typical micro-enterprise has from 2 to 7 employees, has been in operation from 3
months to 10 years (for more advanced countries) and has total assets ranging from
USD 3,000 to 50,000. A small enterprise typically has from 10 to 60 employees, has
been in existence from 1 to 15 years (for more advanced countries) and has total
assets from USD 50,000 to 500,000. The banks working within the EU/EBRD-SME
Facility target micro, small and medium sized enterprises whose size ranges from 1 to
249 employees. On average, banks operating under this programme effectively
8
finance relatively larger clients. At the end of 2003, the average client had 16
employees.
3. EBRD MSME Programmes and Hypothesized Effects
In general, evaluations of providers of finance to the MSME sector have been
limited in their scope to assessments of outreach (measured as the number of
enterprises served), the quality of the loans (typically measured by loan portfolio
arrears ratios), the efficiency of the use of public funds invested (measured as the ratio
of subsidies to the number and volume of loans and as loan officer efficiency
(typically measured as the number of outstanding loans divided by the number of
trained and retrained loan officers in the programme), the sustainability of the
financial intermediaries (full cost return on equity, and of various social objectives
(number of women borrowers, number of clients below poverty line, average loan
balance per borrower in relation to GNI per capita, and regional dispersion of loans).6
EBRD’s internal evaluations of its MSME programmes have been limited to an
assessment of the impact of these programmes on the ability of the banking sector to
provide finance to MSMEs on a sustainable basis. There has been no evaluation of the
impact of these programmes on the enterprises that benefited from the associated bank
finance.
As evidenced by a number of enterprise surveys, access to external sources of
finance remains an important business constraint for small firms in transition
economies. In particular the 2005 Business Environment and Enterprise Performance
Survey (BEEPS) showed that, although access to external finance is becoming with
6 http://www.mixmarket.org/en/demand/demand.profile.comparison.asp
9
time a less severe business constraint, MSMEs in transition economies suffer from
poor access to external finance to a larger degree than MSMEs in mature economies.7
By managing to relieve MSMEs of one of the most frequently quoted (and
most highly rated) constraints to doing business and expansion, we expect that firms
receiving bank finance have an overall better performance and a higher survival rate
than firms that do not manage to access bank loans. We also expect that firms benefit
from bank loans would rate finance as a lesser obstacle to doing business than firms
which do not manage to obtain a bank loan. Moreover, due to the different lending
methodology adopted by the micro-lending programmes and microfinance banks,
which typically results in faster disbursements and relatively easier loan application
procedures due to the more flexible use of collateral, we would expect clients of these
types of programmes to perceive finance as a lesser business obstacle overall than
clients of other banks. Longer maturity bank finance offered by banks under the
EBRD programmes is expected to result in higher investment ratios for their clients
than for other firms.
To test these hypotheses we constructed a questionnaire covering the following
areas of enterprise behaviour:
1. Financial performance (profit, output, sales, exports, investment, and leverage
ratio)
2. Employment dynamics (changes in both full-time and part-time staff)
3. Market expansion (changes in market share, changes in sectors of activity)
4. Relations between firms and financial providers (ease of obtaining external
finance prior to 2002 and after 2002, access to other bank loans)
5. Perception of obstacles to doing business
7 See also Pissarides, Singer and Svejnar (2003) for earlier systematic evidence.
10
4. The Survey, Sample and Basic Statistics
During the first half of 2005 we administered the questionnaire to a sample of
1,272 MSMEs (defined as firms with fewer than 250 employees) in Bulgaria,
Georgia, Ukraine, and Russia.8 In each country, these MSMEs represent a stratified
random sample of enterprises that in 2002 received finance from EBRD’s MSME
financial intermediaries (roughly two thirds of the overall sample per country) and
enterprises that by the time of the survey had not received finance from EBRD
intermediaries but were in existence in 2002 (one third of the overall sample per
country). The former MSMEs represent our treatment group and the latter ones
constitute our control group. The treatment group firms are a random sample stratified
by employment size and sector.9 The treatment group is split into two sub-samples
(roughly equal in size), one of which includes enterprises that received finance from a
microfinance bank in 2002 and the other including enterprises which received finance
in 2002 from a local bank participating either in a micro-lending programme or in the
EU/EBRD SME Facility.
We restricted the scope of the current research to the trade-retail and
manufacturing sectors, but no quotas were applied to the sectors. In practice most
interviewed enterprises were in the trade sector, as the majority of companies which 8 The selection of countries in which the survey was run was based on a number of factors. The first is the number of loans extended by each financial intermediary in the EBRD programme(s) being statistically significant (for statistical purposes this had to be at least 250). Second, to allow for a comparison of the impact on MSMEs of the different quality of finance provided by different types of financial intermediaries, the presence of both dedicated microfinance institution and existing local banks administering targeted credit lines being desirable. Finally, in the case of a large country, the selected regions needing to overlap with regions in which the 2002 BEEPS was run. 9 Except for Ukraine it was not possible to find sufficient enterprises in the last employment category as most of the banks working for EBRD did not extend a sufficient number of loans to this category of enterprises. Also in the case of TUB in Georgia it was impossible to interview the specified quota of 100 enterprises per each bank due to the small number of loans extended by this bank in 2002 combined with business failures and inability to reach the enterprises which benefited from TUB loans. This failure was compensated by adding more enterprises from the Procredit Bank in Georgia. In Bulgaria, the Hebros Bank and Procredit Bank had several inaccurate contact entries and the sample was hence drawn with replacement.
11
borrowed from these banks are in the trade sector. Micro-enterprises constitute the
bulk of the loan portfolio clients of the financial intermediaries used by the EBRD. In
the case of microfinance institutions and micro-lending programmes through
participating banks, micro-enterprises account on average for two-thirds of the
volume and 90 per cent of the number of loans.10 As the role of the micro enterprises
in the financial intermediaries portfolios is so large, this is reflected in the size of
sample strata by size class11. Because we wanted to analyse the impact of EBRD
finance on enterprises of all sizes, we aimed at having all size classes represented in
the sample. Yet, due to total sample size limitations, in some cases the sample
stratification does not necessarily mirror exactly the financial intermediaries portfolio
composition, although it gives a heavier weight to micro-enterprises (54 per cent of
total number of surveyed enterprises) than small (36 per cent) or medium sized
enterprises (10 per cent).
Table 1 shows the sample composition by size class and sector for both
control and treatment groups. The control group firms were selected as a stratified
random sample from marketing lists, internet databases, yellow pages and
interviewers’ walk-ins.
The summary statistics related to the key variables used in our analysis are
provided in Table 1A. As may be seen from the table, the variables have reasonable
values and display considerable variation in within and across countries. Given that
the matching of the control group to the treatment group was structured around
employment size and sector of the firms, other variables than employment show a
10 In the case of the programmes run under the EU/EBRD SME Facility these data is unknown as monitoring of the use of the proceeds of the Facility is based on its sub-loans’ size rather than on its sub-borrowers’ size. 11 Quotas were specified for the size composition of the sample of enterprises to be interviewed (50 per cent of the sample had to employ up to 9 workers, 20 per cent between 10 and 24 workers, 15 per cent between 25 and 49 and 15 per cent between 5 and 249).
12
larger variation. For example, both in Bulgaria and Russia, the presence of some
companies with large revenues in the control groups is evidenced by much larger
mean values for revenues in the control group than in the treatment group.
Loans and Firm Survival and Job Creation
While in the case of the MSME clients of EBRD programmes we know which
firms exited and which survived, the above control group does not include enterprises
that were in existence in 2002 and exited thereafter. We therefore cannot compare the
survival rates of the (EBRD) treatment and the control group of firms. As a result, in
order to carry out such a comparison, we selected as another control group for this
purpose enterprises that were respondents in the 2002 BEEP survey and in 2002
expressed willingness to be re-interviewed in 2005. The response (re-interviewing)
rate was 41% in Bulgaria, 36% in Georgia, 12% in Russia, and 40% in Ukraine. The
reasons for these less than 100% re-interviewing rates most importantly a refusal to
co-operate, followed by firm exit (death). We can distinguish these reasons, and for
the purposes of calculating firm survival, we have a complete count for of the BEEPs
firms. The BEEPs firms that agreed to be re-interviewed in 2005 all answered a
reduced version of the questionnaire covering mainly employment dynamics. In our
analysis, we divide these firms into those that were recipients and non-recipients of
loans and we estimated net job creation of EBRD programmes.
In Table 2 we present the exit (death) rate of screened companies subdivided
into five categories for each country: companies that received a loan from EBRD
owned microfinance banks; companies that received a loan from an EBRD partner
bank; companies in the BEEPS control group; companies in the BEEPS control group
that received a loan; companies in the BEEPS control group that did not receive a
13
loan in 2002. The exit rates over the 2002-05 period (all companies existed in 2002) is
calculated on the basis of ascertained cessation of business. The exit rates of the
control groups, taken as a whole, are significantly higher than those of the enterprises
which benefited from EBRD loans. Exit rates of the part of the control groups that
benefited from a bank loan in 2002 are also consistently higher than exit rates of
SMEs which benefited from an EBRD loan. However, within the control group exit
rates for bank loan recipients are not always lower than exit rates of SMEs that did not
receive bank loans. In Georgia and Ukraine the mortality rate for loan recipients in the
control group is in fact higher than for companies that did not receive a loan. This
raises the possibility that the quality of the loan finance received by enterprises in
these two countries might not have been appropriate for the recipients (e.g., loan
maturity may have been too short or with conditions that did not suit the firm’s
needs).
An interesting question that arises in evaluations of programmes that support
provision of finance to the smallest enterprises is whether these programmes result in
job creation. Many donors, governments and politicians believe, rightly or wrongly,
that supporting SMEs will result in job creation. Their direct and indirect support to
the development of such programmes often has been targeted, even if only implicitly,
to increasing employment. We calculated net job creation rates for firms in both the
treatment group and the control group, and for firms in the control group stratified
according to having received a loan or not. Table 3 contains the net job creation rates
for firms in both treatment and control groups for each country12. Net job creation
rates are positive for firms in the treatment group as a whole, for firms in the
treatment group in each country and for firms which were clients of each financial 12 Data for 2002 employment in the companies that received finance from the EBRD partner bank in Russia had to be approximated, as the participating bank did not provide exact employment figures for this group of companies, but rather provided a range.
14
intermediary used by the EBRD. The net job creation rates are in excess of the net
job creation rates for firms in the control group as a whole in each country but Russia.
In Russia net job creation rates for the treatment group are positive but lower than job
creation rates for the part of the control group that benefited from a non-EBRD loan,
but are higher than the net job creation rates for the part of the control group which
did not benefit from any bank lending. Net job creation rates in the control group are
negative in Bulgaria, Georgia and Ukraine and positive for Russia. Results for Russia
might be biased by the fact that there were not sufficient BEEPS companies that could
be screened in the Nizhny Novgorod area where the treatment group is based. Thus
the Russia BEEPS control group was complemented by BEEPS firms in other regions
which might have experienced faster economic growth rates than Nizhny Novgorod.
Within the control group net job creation rates are higher for the companies that
received a loan than for those who didn’t in Georgia and Russia. The reverse is true in
Bulgaria and Ukraine, which suggests that within the control groups in these two
countries bank lending was associated with investments leading to substitution of
labour with capital (if associated with the effect on exit rates in the same group, this is
particularly true for Bulgaria).
5. The Analytical Framework
Our main goal is to analyze the effects that short- and longer-term loans have
on the performance of MSMEs. In carrying out our analysis, we need to take into
account the fact that our sampled firms differ in terms of whether they received an
EBRD loan in 2002 (treatment versus control group) and also whether and when they
received other loans. In particular, firms in the treatment group may have received
other EBRD or non-EBRD loans before and after 2002, while firms in the control
15
group may have received non-EBRD loans at any time. From an analytical standpoint
there may hence be significant selection problems, with better performing firms for
instance being more able to obtain EBRD and/or non-EBRD loans. If one did not
control for this non-random assignment of firms to loans, one could mistakenly
attribute all of the superior post-2002 performance to loans rather than recognizing
that part may be due to inherently superior performance of the firms that receive
loans. In view of the design of our sample, we strive to control as much as possible
for the treatment and performance of different firms up to 2002, and then focus on
analyzing the impact of subsequent EBRD and non-EBRD loans on performance.
Formally, in the spirit of Ashenfelter and Card (1985), Heckman and Hotz
(1989), and Hanousek, Kocenda and Svejnar (2005), we specify a panel-data
treatment evaluation procedure that fits our context and we supplement it with a set of
instrumental variable estimates. Let Xijt be a given performance indicator, with
subscript i denoting an individual firm with loan of type j, in year t. Moreover, let Lijt
denote loan of type j of firm i in year t (this is a dummy variable having value 1 if
firm received the credit of type j in year t and zero otherwise). A model of
performance may be specified in a logarithmic form as
ijtjijjijjijiijt DtLtXtLtX υϕδγβαα τ ++++++= )()()(ln 11 . (1)
Equation (1) is relatively flexible in that in estimating the performance effect δj of
loans Lijt obtained in the 2002-04 period, the equation allows the 2002-04
performance of firms to reflect all time-invariant differences αι that exist across
individual firms, a possible time-varying effect βj of pre-2002 EBRD and non-EBRD
loans Lij1, a possible time-varying effect γj of the firm’s 2001 (base) year performance
Xij1, and time varying effects ϕ that are specific to individual countries, industries,
16
and years D.13 We take firms that received no loans as the base and their logarithm of
performance is permitted to vary over time at the rate α. For ease of interpretation, the
effect of pre-2002 loans βj is measured relative to α.
Our specification in equation (1) thus controls for the effects on performance
of fixed differences among all firms. It also controls for any linearly time-varying
differences among firms that received or did not receive EBRD or other loans before
2002, inter-firm differences in the initial (2001) performance, country-specific fixed
effects, industry-specific fixed effects (proxying for factors such as the degree of
competition or differences in technology), and annual economy-wide shifts (such as
macro shocks or degree of openness to trade). A particular concern is that we should
ensure that our estimates capture the effect of loans rather than other factors such as
competition. As may be seen from equation (1), we do so by controlling for these
other factors by the firm-specific fixed effects, the effect of initial performance
interacted with the time trend, and the industry-specific and annual time dummy
variables interacted with time.
In specifying equation (1) we allow both current and previous year’s loans to
have an equal effect on current year’s performance. This is an acceptable
approximation since in virtually all cases estimations that allowed the two effects to
be different did not result in significantly different coefficients. The reason for this
lack of difference may be that the accounting information on performance refers to
year-end values, while loans are disbursed throughout the year. In empirical work, we
also allow for two specification of the effect of credit: one where the effect does not
vary with the amount of credit and one where the effect of credit varies with loan size.
13 Any time-invariant effects of these variables (i.e., effects on the level of performance) are captured in αι.
17
For estimating purposes, it is useful to let yijt be the percentage change of Xijt
from t - 1 to t and express equation (1) in the annual rate of change (first-difference)
form as14
ijtjijjijjijijt DLXLy εϕδγβα τ ++∆+++= 11 (2)
where εijt = υijt - υijt-1 is the error term. This specification is more parsimonious and
allows us to estimate all the parameters of interest.
There are three key econometric issues that we need to account for in our
analysis: omitted variables bias, measurement error, and endogeneity of receiving
loans. We address omitted variables bias by including a number of important control
variables that we describe above. In dealing with measurement error in loans,
performance and other variables, we note that the error can induce standard
attenuation as well as more complicated biases in estimated coefficients. As discussed
earlier, in collecting the data set we have placed particular emphasis on identifying
precisely individual loans, as well as carefully collecting several indicators of
performance for the current and preceding periods. We have also checked that there
are no outliers that would seriously affect our estimates.
As to endogeneity of receiving loans, we have already mentioned that there is
a danger that the inherently superior performance of the firms selected for receiving
EBRD or non-EBRD loans could be attributed to loans rather than the non-random
assignment of firms to loans. In the present study, we address this problem as follows.
First, we use the panel data specification in equations (1) and (2) with the
14 Equation (2) may also be viewed as coming from a framework such as that invoked in the endogenous growth literature (e.g., Temple, 1999; Barro and Sala-i-Martin, 1995), where the rate of change of the dependent variable may depend on its initial level (e.g., rate of change of performance being related to an initial level of investment) and some other variables. One may also want to ensure that our estimates capture the effect of loans rather than other factors such as competition. As may be seen from equation (1), we do so by controlling for the extent of competition by the firm-specific fixed effects, the effect of initial performance interacted with the time trend, and the industry-specific and annual time dummy variables interacted with time.
18
aforementioned covariates as a panel data treatment evaluation procedure to control
for the possibility that firms are not assigned to loans at random and that lending
institutions may give loans to firms that are inherently superior or inferior performers.
Second, to deal with what we consider a relatively remote possibility that firms that
received pre-2002 loans from a given source may differ among themselves with
respect to some unobserved characteristics correlated with the rate of change of
performance and not captured adequately by Lij1 and Xij1, we also estimate equation
(2) with firm specific random and fixed effects.15 Finally, since the above approaches
may not fully address all types of endogeneity, especially those where the effect is
time-varying, we also employ an instrumental variable procedure.
Instrumental Variables
We use the Hausman (1978) specification test for assessing endogeneity of the
initial pre-2002 loan status. We employ the first-difference IV method in which we
treat the initial loan status as potentially endogenous and instrument it by IVs that we
describe presently. The test is carried out by differencing the two sets of parameter
estimates and standardizing the vector of differences by the difference in the
covariance matrices of the two sets of estimates. The resulting quadratic form is
asymptotically chi-squared with degrees of freedom equal to the number of
parameters being tested.16
We use the following set of firm-specific instrumental variables that we expect
influence the probability that a firm receives EBRD or non-EBRD loan and can be
excluded from the second stage, rate of change of performance equation: A second
15 In the firm specific fixed effects case, we exclude initial performance and country and industry dummy variables since their effect is captured by the firms specific fixed effects. 16 In practice, some diagonal elements of the covariance matrix are negative. As usual, we carry out the test only for parameters corresponding to the positive diagonal elements, with a corresponding correction to the degrees of freedom, using the generalized inverse matrix (procedure YINVO in TSP 4.5).
19
order polynomial of the distance of the firm from the nearest regional capital (Q7.1),
this distance interacted with time, industry-specific average perception of the
respondents about the availability of financing from banks in a given year (Q10.1 –
industry-specific average perception of the respondents about the corruption of bank
officials in a given year, industry-specific average perception of the respondents about
the number of days it takes to obtain a loan in a given year, dummy variable reflecting
whether the firm adopted international accounting standards at least one year before a
loan is granted, the ratios of full time to part time employees and full time to
temporary employees,17 gender of the CEO and CEO gender interacted with lagged
firm size (total employment). All the instrumental variables pass the Hansen over-
identification test.
6. The Empirical Results
We present our empirical estimates in two parts. First, we discuss the results
related to the effects of EBRD (longer term) and non-EBRD (short-term) loans on
MSME performance. In the second section we examine the determinants of whether a
firm receives a loan and what the loan size is.
The Effects of Loans on Performance
The results that we present in the tables in the text come from OLS estimations
of equation (2). The corresponding random effects estimates are virtually identical.
The fixed effects estimates have greater noise to signal ratio and while generally
yielding similar point estimates of the relevant coefficients, they tend to generate
larger standard errors of the estimates. The random effects, fixed effects and IV
estimates are presented in the tables at the end of the paper.
17 These input ratios can be treated as exogenous under the assumption that the firm is price taker in the input market, minimizes cost and has a Cobb Douglas production technology.
20
In estimating equation (2), we find that the coefficients on pre-2002 loans and
2001 performance are uniformly statistically insignificant. In what follows we hence
report the estimates from specifications that exclude these two variables.
In Table 4 we report the effects of EBRD and non-EBRD loans on investment,
with Panel A providing estimates based on the presence or absence of a loan and
Panel B giving estimates based on the size of the loan relative to revenue. Since in
35% of observations the responding firms report undertaking no investment, we need
to address this issue of zero values in estimating our logarithmic regression. We have
taken several approaches to handling this problem. The results that we report in Table
4 come from specification where we use the annual percentage change in investment
I, calculated as (It - It-1)/[(It + It-1)0.5], with firms reporting zero investment in two
consecutive years being given the value of zero for this percentage change. We have
also estimated a specification of equation (2) in which zero investment is replaced
with country specific median investment level times 0.001. This specification reflects
the assumption that even firms reporting no investment most likely carry out some
minimal investment activities during the year in question. We have tested the
sensitivity of our results to the imputed value of investment and found the results to
be qualitatively similar.18 Finally, we have estimated the change in investment rates
(investment to capital and investment to revenue ratios) and, as a lowerbound
estimate, we have estimated equation (2) only for the firms reporting positive
investment in consecutive years. We do not report all these estimates, but they
indicate that both EBRD and non-EBRD loans have a positive effect on investment
and fixed assets.
18 In particular, we have used replacing the zero investment with country of incorporation mean investment a value time 0.0001 and found the coefficients to be slightly larger but similarly statistically significant.
21
As may be seen from Panel A of Table 4, with the dependent variable being (It
- It-1)/[(It + It-1)0.5], the EBRD loans on average result in a 38 log points (46%)
increase in investment, while non-EBRD loans on average raise investment by 55 log
points (73%). The two estimates are statistically different from each other, implying
that non-EBRD loans have a greater effect on investment that the EBRD ones. The
estimates vary by country, with the greatest investment effects of the presence of
EBRD and non-EBRD loans being reported in Russia (68 log point or 97% and 98 log
points or 166%, respectively), followed by Ukraine (40 log points or 49% and 43 log
points or 54%, respectively) and Bulgaria (25 log points or 28% and 58 log points or
79%, respectively). The estimated EBRD coefficient in Georgia is 23% and
statistically significant, while the coefficient for non-EBRD loans is again positive but
statistically insignificant. The estimates based on the size of the loan, reported in
Panel B of Table 4, indicate that the investment effect of EBRD loans varies with the
size of the loan. The effect is sizable and positive in Bulgaria, Georgia and Russia, but
it is small, negative and only significant at the 10% test level in Ukraine. The effect of
non-EBRD loans on investment does not vary with the size of the loan in Bulgaria
and Russia, but the estimated effect is positive in Georgia and negative in Ukraine.19
19 When we assign firms reporting zero investment the value of 0.001 of the national investment average of the firms that report positive investment, we find a strong positive effect of both EBRD and non-EBRD loans on investment in all four countries (except for non-EBRD loans in Georgia). The estimated effect is in most cases too large to be completely credible, indicating that the 0.001 scaling factor may be too small. The estimates hence provide an upper bound on our estimation. The overall effect across countries 127 log points (256%) for EBRD loans and 169 log points (442%) for non-EBRD loans. The two estimates are statistically different from each other at 7% statistical test level, implying that non-EBRD loans have a greater effect on investment that the EBRD ones. The estimates vary by country, with the greatest effects of the presence of EBRD and non-EBRD loans being reported in Russia (226 log point or 858% and 307 log points or 2054%, respectively), followed by Ukraine (153 log points or 362% and 153 log points or 362%, respectively) and Bulgaria (50 log points or 65% and 127 log points or 256%, respectively). The estimated EBRD coefficient is large in Georgia (98 log points of 166%), but the coefficient for non-EBRD loans, while positive, is statistically insignificant. In the estimates based on the size of the loan, the loan amount is an insignificant determinant of investment growth in the overall regression. In Bulgarian, Georgian and Russian country specific regressions, the EBRD loan size is positively related to performance. In Ukraine the EBRD loan amount is negatively related to investment growth, while in Georgia the larger non-EBRD loan is related to higher investment growth. It is interesting to note that when we run our regressions only on
22
As may be seen from Table 1A, the lack of size effect of non-EBRD loans is not
brought about by a small variance of the loan size within each country.
The estimated effects of loans on fixed assets, based on equation (2), are
reported in Table 5. The estimates of average effects of a loan, reported in Panel A,
are in line with those on investment. In a pooled regression they suggest that an
EBRD loan results in a 10.5% increase and non-EBRD loan in a 14% increase in
fixed assets of the firm. The estimates vary by country, with the greatest average
effects being reported in Bulgaria (15% and 22%, respectively), followed by Ukraine
(12% and 14%) and Russia (10% and 13%). The estimated coefficients in Georgia are
smaller (4% and 5%) and they are statistically insignificant. The estimates based on
the size of the loan return less significant results. When we take into account the size
of the loan in Panel B of Table 5, we show that the effect of the size of an EBRD loan
is statistically insignificant in all countries except for Bulgaria, where the effect is
positive. Hence, the positive effect of loan size that we found with respect to
investment does not translate as readily into the effect on fixed assets. As with
investment, the effect of non-EBRD loan size on fixed assets is only positive in
Georgia. The effect is actually negative in Russia and negative but only marginally
significant in Ukraine.
Overall, the results in Tables 4 and 5 are interesting because they confirm the
anecdotal evidence that firms often use short-term loans for investment purposes,
including investment in fixed assets. Moreover, the coefficient estimates in Panels A
suggest that non-EBRD loans have a somewhat larger effect on investment than
EBRD loans, although the EBRD and non-EBRD loan effects are not statistically
firms reporting positive investment in consecutive years, we still find an overall effect of 24 log points (27 %) for the presence of an EBRD loan and 45 log points (57 %) for the presence of non-EBRD loans.
23
different with respect to fixed assets. The differential effect on investment is
intuitively acceptable in that non-EBRD loans are on average larger than EBRD loans
(see Table 1A). Moreover, EBRD loans are meant to complement (rather than
substitute) other sources of loans.
The results in Panel A of Table 6 indicate that the overall average effects of
EBRD and non-EBRD loans on MSME revenue, measured with data pooled across
the four countries, are positive and statistically significant. Receiving an EBRD loan
on average results in 4.3% higher revenue than would be the case if a firm did not
receive such a loan. The average effect of a non-EBRD loan is estimated at 6.3%. The
two effects are significantly different from zero but not from each other. Estimates by
country indicate that these average effects of EBRD loans are statistically significant
in Georgia (6%), Russia (3%) and Ukraine (12%), while the effect of non-EBRD
loans is significant only in Georgia (13%). The remaining country-specific estimates
are by and large positive but statistically insignificant. The estimates based on the size
of a loan, reported in Panel B, indicate that larger EBRD loans result in lower
revenues than larger loans, especially in Russia and Ukraine. The effect of non-EBRD
loan amount on revenue is only significant in Ukraine and it is negative. Hence, while
loans tend to have a positive effect on firm revenue, the larger loans seem to be “too
large” to have a beneficial impact.
In Table 7 we report estimated effects of loans on labour cost. As may be seen
from Panel A of the table, when measured with data pooled across the four countries
the average effects of both EBRD and non-EBRD loans are positive at 5.1% and 10%,
respectively. Again, the two effects are significantly different from zero but only
significantly different from each other at 6% test level. Estimating these effects by
country indicates that the average effect of EBRD loans on labour cost is statistically
24
significant in Bulgaria (6.4%), Russia (4.8%) and Ukraine (9.5%), while the effect of
non-EBRD loans is significant only in Bulgaria (12.8%) and Ukraine (12.5%). The
remaining country-specific estimates are all positive but statistically insignificant. The
estimates based on the size of a loan, reported in Panel B, indicate that in Ukraine
small EBRD and non-EBRD loans have a positive effect on labour cost but that both
effects decline with the size of the loan. In contrast, in Georgia the effect of non-
EBRD loans varies positively with loan size. The important finding is that most
coefficient on loan size are insignificant, suggesting that the effect of loans on labour
cost does not vary systematically with loan size.
Overall, the estimates in Tables 4-7 indicate that firms use EBRD and non-
EBRD loans to invest and augment both revenues and labour costs. The exceptions
are firms in Bulgaria and Georgia, with the former ones increasing labour cost but not
revenues and the latter ones increasing revenues but not labour costs.
In Table 8 we complement our findings on labour cost by examining the effect
of loans on total employment (results based on full time employment are similar). As
may be seen from Panel A of Table 8, the cross country pooled estimates show that
both EBRD and non-EBRD loans have a positive effect on employment (5.8% and
9.5%, respectively). Again, the two effects are significantly different from zero but
not from each other. Country-specific estimates are positive for both sets of loans in
Bulgaria (13.3% and 11.4%) and Ukraine (8.1% and 24.6%), but they are
insignificant in Georgia and Russia. The labour cost effect of loans in Bulgaria and
Ukraine is hence primarily accounted for by the positive effect of loans on
employment, while in Russia it appears to be primarily driven by the effect of
increased wages (labour cost per worker). The estimates based on the size of a loan,
reported in Panel B, indicate that the employment effect of a larger loan, when
25
significant, is negative. This is the case with EBRD and non-EBRD loans in Georgia,
and also EBRD loans in Bulgaria. It must be noted, however, that the EBRD effects
are only significant at the 10% test level.
Having examined the effect of loans on revenues and costs – the two principal
components of profit -- we next examine the direct effect of loans on profit. As may
be seen from Panel A of Table 9, the cross country pooled estimates indicate that
while EBRD loans on average increase profit by 8%, the effect of non-EBRD loans is
statistically insignificant at 6%.20 Again, in view of the size of the associated standard
errors, the two effects are not significantly different from each other. Country-specific
estimates indicate that the overall effect of EBRD loans on profit is brought about by
a strong effect in Ukraine (13.7%) and to a lesser extent in Georgia (9.2%). The effect
of EBRD loans is statistically insignificant in Bulgaria and Russia, while the effect of
non-EBRD loans is statistically insignificant in all four economies. The estimates
based on the size of a loan, reported in Panel B, indicate that there is a statistically
weak negative effect of EBRD loan size on profit in Ukraine, but that in all other
cases the data suggest that the effect does not vary with the size of the loan. The
relatively frequent insignificance of the effect of loans on profit is not surprising.
MSMEs operate in a competitive setting and it is hence quite likely to see loans result
in an increased scale of operations but not necessarily higher profit. Moreover, as we
have seen in the preceding tables, loans are found to have a positive effect on both
revenues and costs, with the net effect on profit thus being a residual effect. The
limited effect of loans on profitability is also seen when we use profit/revenues as the
dependent variable. In this specification the effect is insignificant in all specifications
except for the effect of EBRD loans in Russia, which is -1.9%. 20 Fewer than 5% of the observations have negative values of profit. These observations have been excluded from estimation. When we estimate the profit equation using the same percent change formula for profit as we did for investment, the results are similar but statistically insignificant.
26
Finally, it must be noted that since the effects of loans on revenues are on the
whole positive, many estimates based on dependent variables expressed in a per
revenue form are statistically insignificant.
Determinants of Loans
The data we collected permit us to address the question of what factors
determine whether a MSME receives an EBRD or non-EBRD loan and what the loan
size is. In Table 10 we present estimates explaining the probability that a firm
receives an EBRD or non-EBRD loan in a given year. The results in Tables 10A and
10B relate to whether the firm receives a loan or not and they are based on a probit
estimation method. The results in Tables 10C and 10D are Tobit estimates because the
variable has a lowerbound at zero. The dependent variable in Tables 10C and 10D is
the ratio of loan size over revenues. The results on Tables 10A and 10C are based on
firms, which had received EBRD loan on 2002 and hence the estimation is done only
for the following years 2003 and 2004.
Results in Table 10A indicate that the firms, which had received an EBRD
loan before 2002 have a higher probability to receive an EBRD loan in 2003 and
2004. In Georgia and Russia we observe a negative effect of having receiving a non-
EBRD credit before 2002 on receiving EBRD credit later on. One possible
explanation is linked to the lack of available information to lenders of borrowers’
credit history. The World Bank’s Credit Information Index21 is 0 (lowest) in Georgia,
Russia and Ukraine and 3 (medium) in Bulgaria. However even in the case of
Bulgaria the public registry cover was extremely poor (2004) at 5/1000 capita and
was almost irrelevant for micro-finance providers as at the time it did not record 21 http://www.doingbusiness.org/ . This index ranges from 0 to 6 and measures the rules affecting the scope, accessibility and quality of credit information available to lenders through either public or private bureaus.
27
information on loans below Leva 10,000 (US$ 6,000). This means that lenders only
have information on their former clients rather than on new clients who may have
borrowed from different lenders. Thus receiving a bank loan does not affect the
probability of receiving a bank loan from a different provider through having gained a
credit history. Lagged change in EBRD credit over years has a positive effect on
receiving credit. Also firms which adopted international accounting standards at least
a year before 2002 are more likely to receive an EBRD loan than firms which did not
adopt such standards. Similar results are observed with respect to receiving non-
EBRD credit (see panel B). Firms which received non-EBRD credit before 2002 are
more likely to receive non-EBRD credit in years 2002-2004. Also older firms, firms
with international accounting standards and firms with male CEO are more likely to
receive non-EBRD credit than young firms, firms which have not adopted
international accounting standards and firms with female CEO.
Results in Tables 10C and 10D indicate that the size of the EBRD loan is
related to the firm’s past credit record and to it having adopted international
accounting standards. The size of non-EBRD loans is dependent on the age of the
firm and the gender of the CEO matters.
7. Conclusions
Our study indicates that bank loans have a significant positive effect on most
performance indicators of micro, small and medium sized enterprises (MSMEs) in the
transition economies. In particular, exit rates related to cessation of business are
higher for companies that did not benefit from EBRD loans (even though they may
have benefited from non-EBRD loans) than for companies that benefited from EBRD
28
loans. Net job creation rates are positive and almost everywhere larger for companies
that benefited from EBRD loans than for companies that did not. Both EBRD and
non-EBRD loans have a positive effect on investment and fixed assets, and they
suggest that in the imperfect capital markets characteristic of the transition economies
firms use even short-term (non-EBRD) loans for investment in fixed capital. The
positive effect of EBRD and non-EBRD loans extends to revenues, labour cost and
employment. Hence, the loans serve the purpose of enabling the MSMEs to expand
production beyond the scale that they could achieve without this source of credit.
Interestingly, the two sets of loans differ in their effect on profitability. The EBRD
loans have a positive effect on profit while the non-EBRD loans have no significant
effect. The relatively limited effect on profitability is intuitively acceptable, given that
firms use the loans to expand both revenues and input use. Moreover, in a competitive
environment, one would expect the effect to translate primarily into expanding scale.
As might be expected, the effect of loans on the performance indicators varies
somewhat across countries and further country-specific analyses will be useful.
Moreover, we find that many of the effects do not vary with the size of the loan and
that some loans may be too big in the sense that they bring about a diminishing return
or even decline in performance. This finding deserves further study as it indicates that
the absorptive capacity of MSMEs may need to be more carefully taken into account.
In terms of determinants of loans, we confirm in the transition economy
context that prior credit (loan) history is an important determinant of the ability of
firms to obtain subsequent loans from the same provider but not from different
providers. The (older) age of the firm, the adoption of international accounting
standards and having a male CEO increase the probability of receiving credit from a
29
non-EBRD provider. In terms of determinants of the size of EBRD loans what matters
is the past credit history and having adopted international accounting standards.
30
References:
Brown, J.D., J. Earle and D. Lup (2002), “Determinants of Small Firm Growth: Finance, Human Capital, Technical Assistance and the Business Environment,” Heriot-Watt University and Central European University mimeo. Morduch, J. (1999), “The Microfinance Promise,” Journal of Economic Literature, 37, 1569 – 1614. Pissarides, F., M Singer and J. Svejnar (2003), “Objectives and constraints of entrepreneurs: evidence from small and medium size enterprises in Russia and Bulgaria,” Journal of Comparative Economics, 31, 503-531.
31 Table 1. Sample size and stratification by size class
Number of Employees Sector Bulgaria 0-9 10-24 25-49 50-249 Trade Industry Service Total
Quota 50 20 15 15 50 40 10 100 Control Group 50 37 20 13 34 41 45 120 Procredit Bank 73 41 17 15 69 46 31 146 Hebros 20 20 7 7 28 22 4 54 Total 143 98 44 35 131 109 80 320
Number of Employees Sector Georgia 0-9 10-24 25-49 50-249 Trade Industry Service Total
Quota 50 20 15 15 50 40 10 100 Control Group 77 21 8 7 54 34 25 113 ProCredit Bank 88 15 6 2 86 16 9 111 TUB 71 17 4 1 53 18 22 93 Total 236 53 18 10 193 68 56 317
Number of Employees Sector Russia 0-9 10-24 25-49 50-249 Trade Industry Service Total
Quota 50 20 15 15 70 20 10 100 Control Group 64 31 11 9 75 25 15 115 KMB 42 39 11 14 71 16 19 106 NBD 64 28 13 9 71 12 31 114 Total 170 98 35 32 217 53 65 335
Number of Employees Sector Ukraine 0-9 10-24 25-49 50-249 Trade Industry Service Total
Quota 50 20 15 15 50 40 10 100 Control Group 47 22 15 16 49 41 10 100 ProCredit Bank 47 23 14 16 52 40 8 100 PrivatBank 50 19 16 15 51 40 9 100 Total 144 64 45 47 152 121 27 300 Total surveyed enterprises 693 313 142 124 693 351 228 1272
32
Table 1A. Average Summary Statistics over 2001-2004
Control group Treatment group
Mean Median St. Dev. Mean Median St. Dev.
Bulgaria
Revenues 1,072 250 3,067 494 221 922 Investment 49 5 178 42 8 120 Fixed assets 224 53 482 178 51 514 Net profits 104 24 283 63 30 148 Labour costs 76 24 190 37 15 73 Total employment 24 12 34 18 9 38 EBRD loan size 0 0 0 39,941 (526) 18,819 72,783 non-EBRD loan size 154,014 (83) 68,669 219,702 50,266 (213) 34,334 62,822 Georgia
Revenues 360 47 1,674 241 89 680 Investment 28 0 149 6 0 16 Fixed assets 210 8 702 50 10 140 Net profits 48 7 172 78 16 351 Labour costs 28 5 113 13 5 27 Total employment 14 4 31 7 3 12 EBRD loan size 0 0 0 19,069 (509) 7,485 36,894 non-EBRD loan size 131,387 (46) 46,456 28,8170 16,368 (251) 8,420 26,407 Russia
Revenues 20,456 2,145 123,056 8,378 3,956 17,081 Investment 706 0 3,148 737 200 2,192 Fixed assets 2,896 237 15,487 2,649 593 8,283 Net profits 1,745 288 4,679 1,994 668 8,092 Labour costs 695 223 1,309 587 300 892 Total employment 19 7 42 17 9 25 EBRD loan size 0 0 0.00 423,388 (559) 197,770 815,225 non-EBRD loan size 1,637,040 (54) 316,431 448,2485 384,107 (311) 200,000 493,348 Ukraine
Revenues 1,723 300 7,558 1,653 120 8,323 Investment 51 0 170 117 9 643 Fixed assets 731 35 6,170 247 24 1,086 Net profits 320 27 1,401 931 28 7,196 Labour costs 82 26 147 145 18 726 Total employment 17 8 23 21 7 39 EBRD loan size 0 0 0 92,777 (450) 22,533 200,665 non-EBRD loan size 108,390 (86) 50,000 184,515 47,384 (242) 15,569 91,805
Note: Figures (except total employment) in Table 1 are in thousands of local currency units adjusted to producer prices.22
22 For all countries the country specific producer price index is used except for Russia for which the Nizhny Novgorod regional producer price index is used.
33 Table 2. Exit rates over 2002-2005
Country Source of credit or control group
Exit rate (in percent)
Total number of screened companies
Partner bank clients 0 156 Microfinance bank clients 1 273 BEEPS Control Group: loan recipients
0 21
BEEPS Control Group: not loan recipients
15 61
Bulgaria
BEEPS Control Group: total
11 82
Partner bank clients 12 212 Microfinance bank clients 13 302 BEEPS Control Group: loan recipients
37 19
BEEPS Control Group: not loan recipients
23 56
Georgia
BEEPS Control Group: total
27 75
Partner bank clients 7 228 Microfinance bank clients 5 306 BEEPS Control Group: loan recipients
0 11
BEEPS Control Group: not loan recipients
22 32
Russia
BEEPS Control Group: total
16 43
Partner bank clients 13 303 Microfinance bank clients 7 208 BEEPS Control Group: loan recipients
35 23
BEEPS Control Group: not loan recipients
16 137
Ukraine
BEEPS Control Group: total
19 160
34 Table 3. Net job creation, 2002-2005 Country Source of credit or
control group Net job creation Total number of
screened companies
Partner bank clients 0.42 156 Microfinance bank clients 0.33 273 Treatment Group total 0.35 BEEPS Control Group: loan recipients
-0.21 21
BEEPS Control Group: not loan recipients
-0.1 61
Bulgaria
BEEPS Control Group: total
-0.13 82
Partner bank clients 0.1 212 Microfinance bank clients 0.06 302 Treatment Group total 0.08 BEEPS Control Group: loan recipients
-0.07 19
BEEPS Control Group: not loan recipients
-0.15 56
Georgia
BEEPS Control Group: total
-0.12 75
Partner bank clients 0.17 228 Microfinance bank clients 0.07 306 Treatment Group total 0.11 BEEPS Control Group: loan recipients
0.61 11
BEEPS Control Group: not loan recipients
0.04 32
Russia
BEEPS Control Group: total
0.34 43
Partner bank clients 0.28 303 Microfinance bank clients 0.11 208 Treatment Group total 0.17 BEEPS Control Group: loan recipients
-0.42 23
BEEPS Control Group: not loan recipients
0 137
Ukraine
BEEPS Control Group: total
0.1 160
35 Table 4A. OLS regression of the effect of loans on investment with loan dummies ALL Bulgaria Georgia Russia Ukraine EBRD loan dummy 0.375 0.247 0.229 0.682 0.397
[0.034]**
* [0.067]**
* [0.081]**
* [0.062]**
* [0.061]**
* Non-EBRD loan dummy 0.547 0.582 0.154 0.976 0.43
[0.063]**
* [0.122]**
* [0.172] [0.132]**
* [0.092]**
* Georgia -0.252
[0.041]**
* Russia -0.154
[0.036]**
* Ukraine -0.123
[0.036]**
* Year 2003 -0.169 -0.006 0.2 -0.496 -0.279
[0.049]**
* [0.101] [0.132] [0.067]**
* [0.086]**
* Year 2004 -0.091 -0.18 0.453 -0.214 -0.339
[0.045]** [0.091]** [0.122]**
* [0.070]**
* [0.064]**
* Constant 0.126 0.176 -0.342 -0.06 0.101 [0.053]** [0.105]* [0.144]** [0.100] [0.060]* Observations 3724 952 894 982 896 R-squared 0.05 0.03 0.02 0.21 0.09 Investment growth is defined as (It-It-1)/0.5(It+It-1). If (It+It-1)=0 then investment growth has assigned 0. All regressions included industry dummies. Robust standard errors in brackets. * significant at 10%; ** significant at 5%; *** significant at 1% Table 4B. OLS regression of the effect of loans on investment with loan size
ALL Bulgaria Georgia Russia Ukraine EBRD loan dummy 0.375 0.094 0.153 0.614 0.416 [0.035]*** [0.079] [0.092]* [0.065]*** [0.063]*** Non-EBRD loan dummy 0.545 0.433 0.115 0.952 0.472 [0.068]*** [0.142]*** [0.184] [0.152]*** [0.103]*** EBRD*loan amount 0.001 0.821 0.444 0.547 -0.034 [0.034] [0.311]*** [0.198]** [0.204]*** [0.018]* Critical level of EBRD loan amount 1219.3 [630.1]* Non-EBRD*loan amount 0.006 0.558 0.112 0.177 -0.103 [0.064] [0.394] [0.040]*** [0.433] [0.041]** Critical level of non-EBRD loan amount 460.25 [137.56]*** Georgia -0.252 [0.041]*** Russia -0.154 [0.036]*** Ukraine -0.123 [0.036]*** Year 2003 -0.169 -0.014 0.188 -0.501 -0.281 [0.050]*** [0.102] [0.131] [0.067]*** [0.086]*** Year 2004 -0.092 -0.199 0.444 -0.223 -0.338 [0.045]** [0.091]** [0.121]*** [0.070]*** [0.064]*** Constant 0.128 0.206 -0.339 -0.054 0.101
36 [0.053]** [0.106]* [0.145]** [0.101] [0.061]* Observations 3718 947 894 982 895 R-squared 0.05 0.05 0.03 0.21 0.09
Investment growth is defined as (It-It-1)/0.5(It+It-1). If (It+It-1)=0 then investment growth has assigned 0. All regressions included industry dummies. Robust standard errors in brackets. * significant at 10%; ** significant at 5%; *** significant at 1%
37 Table 5A. OLS regression of the effect of loans on fixed assets with loan dummies
ALL Bulgaria Georgia Russia Ukraine EBRD loan dummy 0.105 0.154 0.042 0.097 0.123 [0.018]*** [0.038]*** [0.026] [0.030]*** [0.046]*** Non-EBRD loan dummy 0.14 0.219 0.052 0.131 0.14 [0.032]*** [0.067]*** [0.040] [0.059]** [0.069]** Georgia -0.092 [0.022]*** Russia -0.21 [0.023]*** Ukraine -0.145 [0.025]*** Year 2003 -0.018 0.059 -0.038 -0.036 -0.057 [0.023] [0.042] [0.028] [0.033] [0.070] Year 2004 -0.08 0.028 -0.009 -0.109 -0.235 [0.018]*** [0.035] [0.028] [0.035]*** [0.041]*** Constant 0.179 0.075 0.165 -0.052 0.09 [0.028]*** [0.050] [0.069]** [0.036] [0.037]** Observations 3632 893 876 967 896 R-squared 0.04 0.04 0.02 0.04 0.03 All regressions included industry dummies. Robust standard errors in brackets. * significant at 10%; ** significant at 5%; *** significant at 1% Table 5B. OLS regression of the effect of loans on fixed assets with loan size
ALL Bulgaria Georgia Russia Ukraine EBRD loan dummy 0.107 0.093 0.054 0.068 0.134 [0.018]*** [0.045]** [0.027]** [0.032]** [0.049]*** Non-EBRD loan dummy 0.137 0.166 0.025 0.158 0.169 [0.034]*** [0.072]** [0.039] [0.068]** [0.076]** EBRD*loan amount -0.01 0.327 -0.072 0.236 -0.02 [0.009] [0.166]* [0.062] [0.213] [0.011]* Critical level of EBRD loan amount 1071.5 75.3 667.9 [975.2] [64.7] [308.7]** Non-EBRD*loan amount 0.01 0.197 0.074 -0.214 -0.07 [0.034] [0.188] [0.010]*** [0.105]** [0.036]* Critical level of non-EBRD loan amount 73.9 240.4 [25.6]*** [104.2]** Georgia -0.092 [0.022]*** Russia -0.21 [0.023]*** Ukraine -0.144 [0.026]*** Year 2003 -0.018 0.056 -0.037 -0.037 -0.058 [0.023] [0.042] [0.028] [0.033] [0.070] Year 2004 -0.08 0.021 -0.007 -0.112 -0.234 [0.018]*** [0.035] [0.028] [0.035]*** [0.041]*** Constant 0.178 0.087 0.154 -0.05 0.09 [0.028]*** [0.050]* [0.070]** [0.035] [0.038]** Observations 3631 893 876 967 895 R-squared 0.04 0.05 0.02 0.05 0.03
All regressions included industry dummies. Robust standard errors in brackets. * significant at 10%; ** significant at 5%; *** significant at 1%
38 Table 6A. OLS regression of the effect of loans on revenue with loan dummies
ALL Bulgaria Georgia Russia Ukraine EBRD loan dummy 0.043 0.018 0.06 0.032 0.069 [0.012]*** [0.030] [0.024]** [0.014]** [0.026]*** Non-EBRD loan dummy 0.063 0.026 0.128 -0.004 0.09 [0.025]** [0.043] [0.051]** [0.024] [0.057] Georgia 0.004 [0.019] Russia -0.123 [0.016]*** Ukraine -0.037 [0.020]* Year 2003 -0.043 -0.015 -0.031 -0.04 -0.085 [0.014]*** [0.034] [0.027] [0.015]*** [0.034]** Year 2004 -0.076 -0.032 0.016 -0.087 -0.203 [0.012]*** [0.026] [0.025] [0.015]*** [0.028]*** Constant 0.132 0.167 0.185 0.01 0.101 [0.025]*** [0.044]*** [0.075]** [0.031] [0.036]*** Observations 3705 935 894 982 894 R-squared 0.04 0.01 0.05 0.06 0.06
All regressions included industry dummies. Robust standard errors in brackets. * significant at 10%; ** significant at 5%; *** significant at 1% Table 6B. OLS regression of the effect of loans on revenue with loan size
ALL Bulgaria Georgia Russia Ukraine EBRD loan dummy 0.051 0.04 0.082 0.042 0.083 [0.012]*** [0.035] [0.030]*** [0.016]*** [0.027]*** Non-EBRD loan dummy 0.07 0.034 0.124 -0.007 0.118 [0.027]** [0.048] [0.053]** [0.026] [0.064]* EBRD*loan amount -0.028 -0.122 -0.129 -0.083 -0.026 [0.010]*** [0.108] [0.131] [0.044]* [0.007]*** Critical level of EBRD loan amount 178.8 32.9 63.6 50.9 319.6 [68.2]*** [28.9] [53.1] [25.7]** [114.0]*** Non-EBRD*loan amount -0.024 -0.028 0.012 0.026 -0.069 [0.031] [0.071] [0.014] [0.056] [0.022]*** Critical level of non-EBRD loan amount 289.8 119.9 172.1 [340.4] [278.3] [56.4]*** Georgia 0.004 [0.019] Russia -0.124 [0.016]*** Ukraine -0.032 [0.020] Year 2003 -0.042 -0.015 -0.028 -0.04 -0.086 [0.014]*** [0.034] [0.027] [0.015]*** [0.034]** Year 2004 -0.075 -0.029 0.019 -0.086 -0.201 [0.012]*** [0.026] [0.025] [0.015]*** [0.028]*** Constant 0.13 0.164 0.179 0.009 0.099 [0.025]*** [0.043]*** [0.076]** [0.031] [0.036]*** Observations 3705 935 894 982 894 R-squared 0.05 0.02 0.05 0.06 0.07
All regressions included industry dummies. Robust standard errors in brackets. * significant at 10%; ** significant at 5%; *** significant at 1%
39 Table 7A. OLS regression of the effect of loans on labour costs with loan dummies
ALL Bulgaria Georgia Russia Ukraine EBRD loan dummy 0.051 0.064 0.009 0.048 0.095 [0.014]*** [0.027]** [0.028] [0.022]** [0.036]*** Non-EBRD loan dummy 0.1 0.128 0.058 0.062 0.125 [0.025]*** [0.047]*** [0.042] [0.053] [0.050]** Georgia -0.038 [0.020]* Russia -0.077 [0.018]*** Ukraine -0.044 [0.021]** Year 2003 0.007 0.035 -0.034 0.055 -0.044 [0.017] [0.031] [0.031] [0.022]** [0.048] Year 2004 -0.089 -0.042 -0.058 -0.033 -0.24 [0.014]*** [0.029] [0.030]* [0.023] [0.031]*** Constant 0.135 0.142 0.157 0.025 0.11 [0.025]*** [0.038]*** [0.061]** [0.037] [0.037]*** Observations 3460 909 763 965 823 R-squared 0.03 0.03 0.02 0.03 0.07
All regressions included industry dummies. Robust standard errors in brackets. * significant at 10%; ** significant at 5%; *** significant at 1% Table 7B. OLS regression of the effect of loans on labour costs with loan size
ALL Bulgaria Georgia Russia Ukraine EBRD loan dummy 0.055 0.075 0.032 0.051 0.103 [0.015]*** [0.032]** [0.032] [0.024]** [0.039]*** Non-EBRD loan dummy 0.109 0.123 0.038 0.061 0.169 [0.030]*** [0.052]** [0.051] [0.056] [0.056]*** EBRD*loan amount -0.012 -0.062 -0.128 -0.026 -0.012 [0.006]* [0.080] [0.093] [0.055] [0.006]* Critical level of EBRD loan amount 443.5 122.2 24.6 197.8 866.6 [234.1]* [141.2] [21.8] [392.1] [425.8]** Non-EBRD*loan amount -0.032 0.021 0.051 0.01 -0.107 [0.047] [0.073] [0.015]*** [0.149] [0.038]*** Critical level of non-EBRD loan amount 338.2 158.6 [448.8] [54.3]*** Georgia -0.037 [0.020]* Russia -0.077 [0.018]*** Ukraine -0.04 [0.022]* Year 2003 0.007 0.034 -0.032 0.056 -0.047 [0.017] [0.031] [0.031] [0.022]** [0.046] Year 2004 -0.089 -0.041 -0.055 -0.032 -0.242 [0.014]*** [0.029] [0.030]* [0.024] [0.030]*** Constant 0.136 0.141 0.147 0.024 0.113 [0.025]*** [0.038]*** [0.063]** [0.037] [0.038]*** Observations 3459 909 763 965 822 R-squared 0.03 0.03 0.02 0.03 0.08
All regressions included industry dummies. Robust standard errors in brackets. * significant at 10%; ** significant at 5%; *** significant at 1%
40 Table 8A. OLS regression of the effect of loans on employment with loan dummies
ALL Bulgaria Georgia Russia Ukraine EBRD loan dummy 0.058 0.133 -0.007 0.03 0.081 [0.013]*** [0.026]*** [0.023] [0.020] [0.031]** Non-EBRD loan dummy 0.095 0.114 -0.015 0.024 0.246 [0.026]*** [0.038]*** [0.047] [0.037] [0.060]*** Georgia -0.032 [0.017]* Russia -0.014 [0.017] Ukraine 0.061 [0.021]*** Year 2003 0.085 0.383 -0.019 -0.013 -0.045 [0.013]*** [0.025]*** [0.021] [0.023] [0.027]* Year 2004 -0.083 -0.288 -0.024 -0.061 0.049 [0.016]*** [0.034]*** [0.021] [0.025]** [0.038] Constant 0.054 0.032 0.126 0.101 0.049 [0.024]** [0.039] [0.056]** [0.037]*** [0.031] Observations 3335 886 780 908 761 R-squared 0.06 0.42 0.04 0.02 0.06
All regressions included industry dummies. Robust standard errors in brackets. * significant at 10%; ** significant at 5%; *** significant at 1% Table 8B. OLS regression of the effect of loans on employment with loan size
ALL Bulgaria Georgia Russia Ukraine EBRD loan dummy 0.06 0.169 0.018 0.033 0.082 [0.013]*** [0.037]*** [0.027] [0.022] [0.032]** Non-EBRD loan dummy 0.11 0.095 0.033 0.006 0.246 [0.028]*** [0.048]** [0.042] [0.038] [0.064]*** EBRD*loan amount -0.011 -0.212 -0.144 -0.028 -0.005 [0.012] [0.128]* [0.078]* [0.062] [0.008] Critical level of EBRD loan amount 533.7 79.8 12.7 120.0 1751.2 [568.7] [38.3]** [15.9] [244.1] [2772.3] Non-EBRD*loan amount -0.052 0.072 -0.126 0.127 0 [0.053] [0.112] [0.019]*** [0.096] [0.053] Critical level of non-EBRD loan amount 209.2 26.2 [198.3] [32.4] Georgia -0.032 [0.017]* Russia -0.016 [0.017] Ukraine 0.062 [0.021]*** Year 2003 0.085 0.381 -0.014 -0.014 -0.044 [0.013]*** [0.026]*** [0.021] [0.023] [0.027] Year 2004 -0.084 -0.287 -0.021 -0.061 0.05 [0.016]*** [0.035]*** [0.021] [0.025]** [0.038] Constant 0.054 0.03 0.135 0.102 0.046 [0.024]** [0.039] [0.053]** [0.037]*** [0.031] Observations 3330 882 780 908 760 R-squared 0.06 0.42 0.06 0.02 0.06
All regressions included industry dummies. Robust standard errors in brackets. * significant at 10%; ** significant at 5%; *** significant at 1%
41 Table 9A. OLS regression of the effect of loans on net profit with loan dummies
ALL Bulgaria Georgia Russia Ukraine EBRD loan dummy 0.08 0.06 0.092 0.036 0.137 [0.021]*** [0.049] [0.048]* [0.031] [0.044]*** Non-EBRD loan dummy 0.062 0.015 0.162 -0.024 0.126 [0.043] [0.086] [0.105] [0.077] [0.077] Georgia 0.044 [0.032] Russia -0.12 [0.027]*** Ukraine -0.014 [0.030] Year 2003 -0.05 0.02 -0.058 -0.045 -0.137 [0.031] [0.065] [0.064] [0.045] [0.073]* Year 2004 -0.035 -0.015 0.119 -0.138 -0.107 [0.028] [0.058] [0.065]* [0.042]*** [0.059]* Constant 0.09 0.136 0.185 -0.008 0.054 [0.040]** [0.078]* [0.094]* [0.056] [0.055] Observations 3520 899 824 978 819 R-squared 0.02 0.01 0.03 0.02 0.02
All regressions included industry dummies. Robust standard errors in brackets. * significant at 10%; ** significant at 5%; *** significant at 1% Table 9B. OLS regression of the effect of loans on net profit with loan size
ALL Bulgaria Georgia Russia Ukraine EBRD loan dummy 0.087 0.122 0.087 0.052 0.152 [0.021]*** [0.058]** [0.052]* [0.032] [0.045]*** Non-EBRD loan dummy 0.063 0.055 0.164 -0.063 0.125 [0.046] [0.093] [0.119] [0.081] [0.086] EBRD*loan amount -0.031 -0.333 0.032 -0.132 -0.03 [0.017]* [0.215] [0.133] [0.082] [0.015]* Critical level of EBRD loan amount 278.8 36.5 39.3 515.4 [156.8]* [20.1]* [29.7] [275.2]* nonEBRD*loan amount -0.003 -0.144 -0.015 0.296 0.003 [0.038] [0.120] [0.468] [0.350] [0.043] Critical level of non-EBRD loan amount 1951.6 38.2 1068.3 21.2 [22308.0] [60.1] [32114.8] [29.3] Georgia 0.044 [0.032] Russia -0.12 [0.027]*** Ukraine -0.009 [0.030] Year 2003 -0.049 0.022 -0.059 -0.045 -0.137 [0.031] [0.065] [0.064] [0.045] [0.073]* Year 2004 -0.034 -0.009 0.119 -0.137 -0.106 [0.028] [0.057] [0.065]* [0.042]*** [0.059]* Constant 0.087 0.126 0.186 -0.008 0.052 [0.040]** [0.077] [0.095]* [0.055] [0.056] Observations 3520 899 824 978 819 R-squared 0.02 0.01 0.03 0.03 0.02
All regressions included industry dummies. Robust standard errors in brackets. * significant at 10%; ** significant at 5%; *** significant at 1%
42 Table 10A. Probit regression of the effect of firm characteristics on receiving EBRD credit (marginal effects)
ALL Bulgaria Georgia Russia Ukraine Year 2003 -0.03 -0.094 -0.18 0.038 0.072 [0.032] [0.057]* [0.085]** [0.077] [0.051] Georgia -0.158 [0.045]*** Russia -0.132 [0.044]*** Ukraine -0.146 [0.047]*** EBRD credit before 2002 0.156 0.102 0.235 0.235 0.076 [0.029]*** [0.062] [0.064]*** [0.055]*** [0.067] non-EBRD credit before 2002 -0.083 0.073 -0.432 -0.425 0.201 [0.068] [0.080] [0.161]*** [0.107]*** [0.165] Lagged change in EBRD credit 0.1 0.133 0.169 0.069 0.061 [0.023]*** [0.045]*** [0.060]*** [0.056] [0.039] Lagged change in non-EBRD credit -0.118 0.114 -0.641 -0.232 -0.088 [0.069]* [0.104] [0.142]*** [0.158] [0.138] Age 0.004 -0.019 0.011 -0.001 0.008 [0.006] [0.037] [0.010] [0.037] [0.039] Age squared 0 0.001 0 0 0 [0.000] [0.002] [0.000]* [0.002] [0.002] Distance from regional capital 0.002 -0.002 0.004 [0.002] [0.005] [0.002] Distance squared 0 0 0 [0.000] [0.000] [0.000] Established at 2002 0.06 0.12 0.004 0.097 [0.068] [0.132] [0.093] [0.214] International accounting standards 0.311 0.339 0.267 [0.037]*** [0.062]*** [0.124]** Woman CEO 0.021 0.086 0.094 -0.034 -0.046 [0.030] [0.058] [0.065] [0.065] [0.070] Woman CEO interacted with total employment -0.001 -0.002 -0.009 -0.002 0.003 [0.001] [0.002] [0.008] [0.003] [0.002] Observations 1648 397 406 438 377
All regressions included industry dummies. Robust standard errors in brackets. * significant at 10%; ** significant at 5%; *** significant at 1%
43 Table 10B. Probit regression of the effect of firm characteristics on receiving non-EBRD credit (marginal effects) ALL Bulgaria Georgia Russia Ukraine Year 2003 0.058 0.035 0.108 0.053 0.058 [0.011]*** [0.021]* [0.035]*** [0.020]*** [0.023]** Year 2004 0.06 0.051 0.066 0.082 0.059 [0.010]*** [0.019]*** [0.025]*** [0.026]*** [0.026]** Georgia -0.022 [0.010]** Russia 0.002 [0.011] Ukraine 0.003 [0.012] EBRD credit before 2002 -0.039 -0.074 -0.009 -0.046 -0.036 [0.007]*** [0.015]*** [0.008] [0.013]*** [0.021]* non-EBRD credit before 2002 0.236 0.135 0.248 0.244 0.497 [0.037]*** [0.046]*** [0.103]** [0.097]** [0.091]*** Lagged change in EBRD credit -0.016 -0.032 -0.013 -0.004 -0.009 [0.007]** [0.017]* [0.006]** [0.013] [0.017] Lagged change in non-EBRD credit 0.068 0.042 0.045 0.055 0.126 [0.012]*** [0.025]* [0.017]*** [0.020]*** [0.030]*** Age 0.002 0.006 0.001 0.001 0.011 [0.001]* [0.004]* [0.001]* [0.006] [0.012] Age squared 0 0 0 0 -0.001 [0.000] [0.000] [0.000] [0.000] [0.001] Distance from regional capital 0 -0.001 0 [0.000] [0.002] [0.001] Distance squared 0 0 0 [0.000] [0.000] [0.000] Established at 2002 -0.005 0.053 -0.008 0 [0.017] [0.115] [0.009] [0.027] International accounting standards 0.199 0.463 0.162 0.289 0.06 [0.050]*** [0.204]** [0.055]*** [0.155]* [0.065] Woman CEO -0.021 -0.033 0.003 -0.022 -0.038 [0.008]*** [0.018]* [0.009] [0.013]* [0.020]* Woman CEO interacted with total employment 0 0 0 0.001 0 [0.000]*** [0.000]*** [0.000] [0.000]*** [0.000] Observations 3815 936 893 954 868 All regressions included industry dummies. Robust standard errors in brackets. significant at 10%; ** significant at 5%; *** significant at 1%
44 Table 10C. Tobit regression of the effect of firm characteristics on EBRD loan size (marginal effects) ALL Bulgaria Georgia Russia Ukraine Year 2003 -0.075 -0.064 -0.048 0 0.078 [0.060] [0.031]** [0.054] [0.036] [0.207] Georgia -0.118 [0.077] Russia -0.145 [0.074]** Ukraine 0.086 [0.077] EBRD credit before 2002 0.239 0.044 0.183 0.084 0.227 [0.056]*** [0.030] [0.043]*** [0.028]*** [0.226] non-EBRD credit before 2002 -0.089 0.056 -0.363 -0.237 0.011 [0.112] [0.040] [0.154]** [0.073]*** [0.589] Lagged change in EBRD credit 0.141 0.058 0.046 0.022 0.234 [0.048]*** [0.026]** [0.041] [0.028] [0.178] Lagged change in non-EBRD credit -0.089 0.028 -0.495 -0.117 0.194 [0.137] [0.053] [0.165]*** [0.091] [0.553] Age 0.012 -0.008 0.003 0.021 0.214 [0.012] [0.018] [0.007] [0.017] [0.134] Age squared 0 0.001 0 -0.001 -0.013 [0.000] [0.001] [0.000] [0.001] [0.008] Distance from regional capital 0.01 0 0.022
[0.003]*** [0.002] [0.008]*
** Distance squared 0 0 0
[0.000]** [0.000] [0.000]*
* Established at 2002 0.089 0.106 0.002 0.017 1.02 [0.122] [0.082] [0.062] [0.085] [0.949] International accounting standards 0.303 -0.047 0.105 0.425 0.74 [0.110]*** [0.143] [0.054]* [0.095]*** [0.431]* Woman CEO 0.088 0.067 0.086 -0.033 0.183 [0.055] [0.029]** [0.042]** [0.030] [0.230] Woman CEO interacted with total employment -0.002 0 0.001 0 -0.002 [0.002] [0.000] [0.004] [0.001] [0.007] Constant -0.275 0.011 -0.13 -0.109 -1.307
[0.120]** [0.080] [0.091] [0.080] [0.517]*
* Observations 1648 400 408 440 400 All regressions included industry dummies. * significant at 10%; ** significant at 5%; *** significant at 1%
45 Table 10D. Tobit regression of the effect of firm characteristics on non-EBRD loan size (marginal effects)
All regressions included industry dummies. * significant at 10%; ** significant at 5%; *** significant at 1%
ALL Bulgaria Georgia Russia Ukraine Year 2003 0.438 0.188 1.38 0.175 0.377 [0.091]*** [0.103]* [0.368]*** [0.062]*** [0.208]* Year 2004 0.446 0.218 0.789 0.222 0.437 [0.088]*** [0.098]** [0.349]** [0.061]*** [0.207]** Georgia -0.169 [0.100]* Russia -0.023 [0.092] Ukraine 0.075 [0.091] EBRD credit before 2002 -0.465 -0.534 -0.286 -0.233 -0.479 [0.100]*** [0.147]*** [0.283] [0.078]*** [0.226]** non-EBRD credit before 2002 0.871 0.371 1.281 0.262 1.598 [0.086]*** [0.090]*** [0.334]*** [0.063]*** [0.220]*** Lagged change in EBRD credit -0.11 -0.21 -0.404 -0.031 0.118 [0.072] [0.091]** [0.235]* [0.046] [0.168] Lagged change in non-EBRD credit 0.546 0.113 1.043 0.19 1.015 [0.095]*** [0.099] [0.305]*** [0.064]*** [0.261]*** Age 0.025 0.021 0.061 0.011 0.184 [0.008]*** [0.016] [0.016]*** [0.025] [0.103]* Age squared 0 0 -0.001 -0.001 -0.013 [0.000]*** [0.000] [0.000]*** [0.002] [0.007]* Distance from regional capital 0.005 -0.003 0.006 [0.003]* [0.009] [0.004] Distance squared 0 0 0 [0.000] [0.000] [0.000] Established at 2002 -0.031 0.25 -0.104 0.007 -5.164 [0.162] [0.240] [0.363] [0.094] [0.000] International accounting standards 0.684 0.649 1.192 0.358 0.191 [0.133]*** [0.286]** [0.281]*** [0.112]*** [0.388] Woman CEO -0.156 -0.109 0.05 -0.074 -0.163 [0.074]** [0.092] [0.231] [0.050] [0.187] Woman CEO interacted with total employment 0.002 0.001 -0.013 0 0.001 [0.001] [0.001] [0.006]** [0.001] [0.007] Constant -1.688 -0.87 -2.8 -0.448 -2.586 [0.147]*** [0.170]*** [0.516]*** [0.105]*** [0.407]*** Observations 3809 955 950 1005 899