WPS6030


Policy Research Working Paper                         6030




            Funding vs. Real Economy Shock
                 The Impact of the 2007–2009 Crisis
                  on Small Firms’ Credit Availability

                                   Gunhild Berg
                               Karolin Kirschenmann




The World Bank
Africa Region
Finance and Private Sector Development Unit
April 2012
Policy Research Working Paper 6030


  Abstract
  This paper analyzes the impact of two distinct shocks                             contrasted with a real economy shock that hit exactly
  stemming from the cross-border transmission of the                                when the bank’s funding position strengthened again.
  2007-2009 crisis on credit availability for small firms.                          The paper finds that during the funding shock (potential)
  The paper uses data from AccessBank Azerbaijan which                              borrowers are discouraged from applying for loans.
  was affected in its liquidity position during the second                          However, for those applications made, the likelihood of
  and third quarters of 2008 by delays in its refinancing.                          loan approval is not affected. The real economy shock, in
  The Azeri real economy was hit by the global crisis from                          contrast, reduces the approval likelihood for SME loans
  the fourth quarter of 2008 onwards with a combined                                in particular, while agro and micro loans are considerably
  decline in oil prices, exports, remittances, and domestic                         less affected. Finally, bank relationships increase credit
  demand. Therefore, a pure supply side shock con be                                availability in good as well as in bad times.




  This paper is a product of the Finance and Private Sector Development Unit, Africa Region. It is part of a larger effort by
  the World Bank to provide open access to its research and make a contribution to development policy discussions around
  the world. Policy Research Working Papers are also posted on the Web at http://econ.worldbank.org. The author may be
  contacted at gberg@worldbank.org.




         The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development
         issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the
         names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those
         of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and
         its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent.


                                                       Produced by the Research Support Team
   Funding vs. real economy shock: The impact of the 2007-2009
                     crisis on small firms’ credit availability

                           Gunhild Berg* and Karolin Kirschenmann**




Keywords: financial crisis, credit availability, emerging markets, SME finance
JEL classification: G01, G15, G21
Sector board: Financial Sector
* Gunhild Berg, The World Bank, Africa Finance and Private Sector Development, 1818 H Street,
NW, Washington, DC 20433/USA, E-Mail: gberg@worldbank.org. ** Karolin Kirschenmann, Aalto
University School of Economics, Department of Finance, Runeberginkatu 22-24, 00100
Helsinki/Finland, E-Mail: karolin.kirschenmann@aalto.fi.

Acknowledgements: We thank Kristina Czura, Ralph De Haas, Gerald Hübner, Michael Jainzik,
Steven Ongena and Eva Terberger, seminar participants at the EBRD as well as participants at the
PEGNet Conference 2010 in Johannesburg and the 2nd European Microfinance Research Conference
in Groningen for helpful comments. We are very grateful to the staff of AccessBank Azerbaijan, in
particular Andrew Pospielovsky and Rovshan Rustamov, among others, who not only provided us
with the data but also gave us valuable insights into the business of the bank and the banking sector in
Azerbaijan. The findings, interpretations, and conclusions are entirely those of the authors and do not
necessarily represent the view of the World Bank, its Executive Directors, or the countries they
represent. All remaining errors are ours.
1 Introduction
     The integration and development of emerging countries’ financial systems benefitted
them by increased supply of finance and economic growth (e.g. Guiso, Sapienza and Zingales
(2004), Bekaert, Harvey and Lundblad (2005) and Giannetti and Ongena (2009)). At the same
time, these countries became more likely to feel the spillovers of shocks, such as the 2007-
2009 financial crisis, that originated unrelated to local economic conditions.1 Banks’ lending
to micro, small and medium enterprises (MSME banks) in emerging countries and their
customers were mostly resilient to the effects of previous financial crises, whereas they are
more likely to be affected by the recent crisis because they have also become more globally
integrated in the past decade (Di Bella (2011), Galema, Lensink and Spierdijk (2011)). The
aim of this paper is to provide evidence on the different channels through which the recent
crisis was transmitted to banks providing lending to MSMEs and their borrowers in emerging
markets.
     The impact of financial crises on the loan portfolio of MSME banks is ex ante unclear.
Banks facing liquidity shortages may be forced to cut down lending. At the same time banks
might be reluctant to reject loan applications of existing borrowers even in times of crisis
because they have invested in costly information acquisition (Rajan (1992), De Haas and Van
Horen (2010)). With respect to the clients, a larger proportion of microenterprises, in contrast
to SME businesses, is engaged in the supply of essential goods and services, for which
demand fluctuates less even in times of crisis (Littlefield and Kneiding (2009)). However, it is
also possible that micro borrowers are more severely hit by the crisis since especially in
advanced and competitive (microfinance) markets the boundaries between micro loans and
consumer finance have more and more blurred (e.g. Christen (2001) and Littlefield and
Rosenberg (2004)). In such markets, many micro entrepreneurs have borrowed from multiple
sources and accumulated high levels of debt which makes them vulnerable to even small
changes in their income. While there is substantial evidence on the effects of the crisis on
bank lending in general,2 the evidence on how MSME banks and their clients are hit is
anecdotal so far. Also, there is only limited evidence on how exactly banks reacted to
different shocks that occurred during the crisis.


1
  The unfolding of the crisis events is e.g. described in Brunnermeier (2009) and Ivashina and Scharfstein
(2010). Popov and Udell (2011), De Haas, Korniyenko, Loukoianova and Pivovarsky (2012) and Ongena,
Peydro and Van Horen (2012) study how borrowers in Emerging Europe are affected due to the cross-border
transmission of the crisis in foreign-owned and internationally funded banks.
2
  De Haas and Van Horen (2010), Huang (2010), Puri, Rocholl and Steffen (2010), De Haas and Van Horen
(2011), De Haas and Van Lelyveld (2011) and Jimenez, Ongena, Peydro and Saurina (2012) provide evidence on
the worldwide effects of the crisis and the transmission mechanisms.

                                                                                                        2
     In this paper, we study two distinct shocks stemming from the global transmission of the
crisis and their respective impact on credit availability for micro, small and medium firms.
We analyze a unique dataset including information from all business loan applications and
loan contracts of AccessBank Azerbaijan between 2006 and 2009. The bank provides an
interesting object of study since it is foreign-owned and largely refinanced in foreign currency
as are many MSME banks in the region. Due to the distortions in international capital
markets, which, for instance, made the placement of a CDO impossible, the bank was hit by
unforeseen delays in its refinancing pipeline in the second and third quarters of 2008 (the first
shock) although its financial performance remained very strong.
     At the same time, Azerbaijan’s economy proved resilient to the international turmoil up
to the last quarter of 2008 when economic growth eventually slowed down (IMF (2008),
Hübner and Jainzik (2009)). The timeline of events in Azerbaijan allows us to study a pure
supply side shock, the Funding shock, while firms had not yet been affected by the economic
crisis (see also Dooley and Hutchison (2009) for this so-called decoupling of emerging
markets in the beginning of the crisis). Then, we can contrast its impact on credit availability
for firms of different sizes and industries from the impact of the Real economy shock (the
second shock). Important for our study is that AccessBank Azerbaijan faced a short period of
tight refinancing at a time when the economic crisis was still absent from the country, while
its refinancing pipeline strengthened again from the fourth quarter of 2008 onwards, i.e.
exactly at the time when Azerbaijan’s economy started to slow down. This setting allows us
to study how banks react to different shocks coming from the funding vs. the real economy
side of their business and the mechanisms they have at hand.
     Observing loan applications and the bank’s decision whether to approve the application,
we assess which loan types (agro, micro and SME) are affected most within the two channels
identified. During the economic slowdown, we expect SME loans to be affected most as their
credit quality is likely to deteriorate most. SME businesses, compared to microenterprises, are
involved in a broader variety of activities and are more likely to provide non-subsistence
goods and services (such as furniture and household appliances or real estate services), for
which demand typically decreases in economic downturns, or may be suppliers to the oil
industry.3 Finally, we examine whether previous bank-borrower relationships help to mitigate
credit constraints.



3
   See e.g. http://www.cdcdevelopmentsolutions.org/eastern-europe-and-caucasus#Azerbijan for efforts to
increase SME capacities to win contracts with the oil industry in Azerbaijan. See IFC (2009) for a comparison of
sectors in which SMEs and individual entrepreneurs are active.

                                                                                                              3
      For the first channel, we study aggregate numbers of loan requests and approvals. Our
results suggest that the bank discourages potential borrowers from applying for new loans
and, to a lesser extent, existing customers from requesting additional loans during the period
of tight refinancing in the second and third quarters of 2008 (see Brown, Ongena, Popov and
Yesin (2011) and Popov and Udell (2011) for the importance of discouragement in Eastern
Europe). The bank’s temporary liquidity squeeze therefore resulted in a slowdown of
portfolio growth because lending had to be limited, thereby affecting all borrower groups.
Generally, one would expect a pure supply shock to mainly influence the bank’s ability to
meet existing loan requests. However, it may also mirror itself in the number of loan
applications because of the particular marketing strategy of AccessBank (and similar MSME
banks)4. MSMEs in Azerbaijan, and those that have never had a loan before in particular, do
not seem to know much about the services banks offer (IFC (2009)). To attract new
borrowers, the bank therefore regularly organizes local community events to market its
products. Also, the bank’s staff may visit the surrounding potential clients and call their
attention to the bank’s business. During the time of tight refinancing, loan officers were
encouraged to take vacation which is equal to curtailing the active marketing of the products
and thus the number of loan applications may also decline due to the temporary squeeze in the
bank’s liquidity.
     Second, we measure credit availability by the likelihood that a loan application is
approved. For the funding shock we find that approval rates for those (potential) borrowers
who actually apply for a loan are not negatively affected in any borrower group. This
corroborates our conclusion that the funding shock mainly affects credit availability via the
discouragement of borrowers. Approval rates remain unaffected by this shock because the
economy was still growing and borrower quality therefore not systematically changed. For the
real economy shock, by contrast, our findings suggest that credit availability depends on the
loan type. For agro loan borrowers, loan approval is hardly affected by this event. This is
surprising on first sight since agro loans are regularly classified as particularly risky because
of their highly correlated risks in case of natural disasters or commodity price fluctuations
(e.g. Wenner, Navajas, Trivelli and Tarazona (2007)). In this economic slowdown, however,
the agro businesses in Azerbaijan remain comparatively unaffected because they mostly grow
fruits or vegetables or raise sheep for the local market. Thus, such loans may offer some


4
  Prominent examples of other MSME banks which use pro-active strategies to distribute their products and
services are the ProCredit Banks, BRI Indonesia and ASA Bangladesh. See also Bankakademie (2000). Most
active marketing aims at increasing financial literacy or raising awareness of savings products, however these
campaigns naturally also bring other banking services to the attention of potential clients.

                                                                                                            4
stability to a bank’s loan portfolio in this kind of global financial and economic crises. The
same seems to apply to the diversification into different loan sizes: the (repeat) micro loans in
our sample show a very small reduction in approval rates during the real economy shock
period which may be explained by their lower risk because they produce subsistence goods
for the local market. SME borrowers who are more likely to encounter fluctuating demand for
their products therefore have to face the greatest cuts to their credit availability. 5
     To confirm that the two crisis shocks have differential effects, we study an exogenous
variation in the severity of the crisis shock and its impact on loan approval. We distinguish
between locations in which the oil and gas industry is most important and the rest of the
country and find that the real economy shock affects credit availability negatively especially
in those locations that can be expected to be hit hardest, i.e. the petroleum locations. The
funding shock, on the contrary, has no negative and no differential (except for SME
borrowers) effect for petroleum- vs. non-petroleum locations.
     Thus, the funding and the real economy shock have distinct impacts on the availability of
credit to agro, micro and SME borrowers. The funding shock leads to a temporary tightening
in credit availability for all borrower groups via the discouragement channel, which is an easy
and fast way to slow down lending activities to save scarce funds. In contrast, the impact of
the economic downturn on credit availability strongly depends on firm size and on the
industry a firm operates in, by mainly reducing the availability of SME loans. This can be
explained by the respectively different deterioration in credit quality due to the worsening
economic situation, while borrower quality was not affected during the time of tight
refinancing. Thus, in the real shock situation a more sophisticated mechanism, i.e. increased
screening effort to single out the less risky clients, seems the more adequate strategy. Finally,
we find that bank-borrower relationships are an important determinant for increasing credit
availability in good as in bad times.
     Our study contributes to the literature on the transmission of the recent crisis and its
impact on bank lending. De Haas and Van Horen (2010) analyze in the syndicated loan
market how banks adjust their lending behavior during a financial crisis and find that the
reduction in bank lending during the crisis can at least partly be attributed to banks’ increased
monitoring and screening efforts. Jimenez, Ongena, Peydro and Saurina (2012) and Puri,


5
  Although SMEs saw their sales drop, they may still have had a need for working capital to bridge temporary
liquidity gaps. Demand for such loans may actually have increased during the economic downturn because, in
contrast to AccessBank, most other Azeri banks faced liquidity problems during the last quarter of 2008. We find
evidence in the empirical analysis that loan applications of new SME borrowers indeed increase after the onset
of the economic slowdown, while loan applications from repeat SME borrowers decrease. However, for both
groups the approval likelihood decreases indicating an increased riskiness of SME borrowers in general.

                                                                                                              5
Rocholl and Steffen (2010) match information from loan applications and loan contracts to
disentangle whether reduced credit availability is the result of demand or supply determinants.
Jimenez, Ongena, Peydro and Saurina (2012) find for their sample of Spanish firms that bank
balance-sheet strength determines loan approval in crisis times. Firm balance-sheet strength
determines loan granting in good as well as in crisis times but matters more during the latter.
Puri, Rocholl and Steffen (2010) study the effects of the financial crisis on retail lending at
German savings banks and find a general decrease of loan demand after the beginning of the
crisis. On the supply side, it turns out that affected banks are less likely to grant loans but that
bank relationships help to mitigate this effect. Finally, Popov and Udell (2011) find cross-
country evidence for Eastern Europe that firms located in areas served by foreign banks
whose parent banks experience financial distress have a higher probability to be credit
constrained.
     In contrast to these papers, our study focuses on two distinct shocks occurring during the
crisis and can thereby disentangle how different shocks impact on different borrower groups.6
Our results imply that a funding shock may also show up in a reduced number of loan
applications (i.e. on the demand side), while a real economy shock may impact on loan
approval (i.e. on the supply side). Undoubtedly, having information on loan applications is an
important feature in better disentangling demand and supply effects in bank lending during
crisis times. Our results, however, indicate that information from loan applications may not be
the remedy per se and future research in this area may need to more explicitly account for the
different mechanisms banks have at hand to react to certain shocks. Our paper sheds some
first light on these mechanisms.
     Finally, our study is related to the literature on the performance of MSME banks and
their portfolios during the East Asian financial crisis of 1997-1998. By analyzing the effects
of the crisis on Bank Rakyat Indonesia (BRI), one of the largest MSME banks in the world,
Patten, Rosengard and Johnston (2001) find nearly no effects on repayment behavior in the
microfinance portfolio, while non-performance rates increased considerably in the SME
portfolio suggesting that micro loans are less risky in times of crises compared to SME loans.
Since higher default rates and reduced credit availability may both be outcomes of higher
credit risk, we complement these results by showing that credit availability is similarly less
affected for micro loans than for SME loans during the 2007-2009 financial and economic
crisis.


6
 In such, it is also related to the studies on specific liquidity or supply shocks to banks (e.g. Khwaja and Mian
(2008) and Paravisini (2008)).

                                                                                                                    6
     The rest of the paper is organized as follows. Section 2 provides information on
Azerbaijan’s economy and how it was affected by the financial crisis. Section 3 describes the
institutional background as well as the data and methodology while section 4 reports the
empirical results. Section 5 concludes.


2 Azerbaijan’s economy and the financial crisis
     Azerbaijan’s economy is highly dependent on the oil and gas sector. The high real
growth rates of 12 percent on average since 1998 with a peak in 2007 of 25 percent would not
have been possible without the oil sector which accounts for about 60 percent of GDP and 95
percent of all exports (Economist Intelligence Unit (2009b)). Furthermore, growth is
concentrated in urban areas and in the extractive sectors leading to high disparities within the
country. Thus, while the average per capita income (in PPP) was at 11,413 USD in 2008 –
corresponding de facto to a middle income country – an estimated 20 percent of the whole
population still lives in poverty (Economist Intelligence Unit (2009b)). The omnipresent
corruption and state interventions in the economy also hinder the future development of the
country.7
     The banking sector in Azerbaijan has shown high growth rates over the last years,
starting from a relatively low level. Total sector assets grew by 78 percent to 8 billion USD in
2007 while the total loan portfolio increased by 102 percent to 5.4 billion USD (Central Bank
Azerbaijan (2008)). Yet, financial intermediation measured as total banking assets over GDP
was with 27 percent at year end 2007 still low.8 The sector remains highly concentrated with
the only state-owned bank (International Bank of Azerbaijan) accounting for 39 percent of
total banking assets in April 2009. The microfinance sector is targeted by 12 banks and 20
non-banks and even though the high number of banks may lead to the impression that the
country is over-banked, access to credit for micro and small enterprises remains one of the
main impediments for further growth especially in rural areas (IFC (2009)).
     The global financial and economic crisis affected Azerbaijan’s economy from the fourth
quarter of 2008 onwards and growth slowed down to 9% in 2009. Compared to other CIS
countries such as Georgia, the Ukraine or Russia the overall macroeconomic and financial
situation seems rather comfortable due to the high growth rates and the constant inflow of
hard currency during the years before the crisis. While other CIS countries suffered from high


7
  In 2008, Azerbaijan has ranked 158th out of 180 countries on the list of Transparency International
(Transparency International (2008)).
8
  For means of comparison, Georgia and Russia have financial intermediation ratios of 42 and 52 percent,
respectively.

                                                                                                      7
currency devaluation, Azerbaijan’s local currency, the Manat, remained stable. However,
what was felt in the economy and in particular in sectors such as trade and construction was
the drop in oil prices. Sectors that remained more or less unaffected were those which are
mainly independent from international markets such as agriculture. Thus, the vulnerability of
the Azerbaijani economy mainly stems from the lack of diversification which will remain a
challenge for the future (Hübner and Jainzik (2009)).
     With respect to the banking sector, indeed most banks had to stop lending at some point
and the quality of the loan portfolio deteriorated from 2.2 to 3.2 percent of loans being
delinquent by more than 90 days in the first quarter of 2009 (Central Bank Azerbaijan
(2009)). With respect to deposits, crisis-effects were visible as business clients increasingly
withdrew their savings to keep their businesses going and households attempted to convert
their savings from Manat to USD due to their fear that the Manat would devaluate. The
Central Bank provided comprehensive stabilization measures in the third quarter of 2008
which included, among others, an emergency facility for liquidity support and a decrease in
the refinancing interest rate from 12 to 3 percent (Hübner and Jainzik (2009)). And while the
low financial intermediation would otherwise be considered unsatisfactory, it helped the
banking sector to remain somewhat immune against the effects of the crisis as the sector as a
whole is only integrated into global financial markets in a limited way. Nevertheless, those
few banks which are globally connected directly felt the effects of the turmoil in financial
markets. Apart from that, all Azeri banks had to deal with an increase in their customers’ risks
when the economic crisis hit the country.


3 Institutional background, data and methodology
3.1 The bank
     The data we use for the empirical analysis was generated using the Management
Information System (MIS) of AccessBank Azerbaijan.9 The mission of AccessBank is to
provide financial services at European standards to micro and small businesses and low and
medium income families while also offering products for larger enterprises.10 AccessBank
was founded by international financial institutions, which are still shareholders today, in
October 2002 under the name Micro Finance Bank of Azerbaijan (MFBA). Until today,




9
  To get a better understanding of the bank’s business, the different loan types and borrower groups as well as the
crisis events, we conducted informal interviews with the bank’s management.
10
   See http://www.accessbank.az/en/index.html or AccessBank (2008) for detailed information on the bank and
its business activity.

                                                                                                                 8
AccessBank is the only fully-fledged bank targeting the micro and small business sector in
Azerbaijan.
    Besides the central branch in Baku, AccessBank has 22 branches located all over the
country both in rural and urban areas. Still, as of July 2009, the loan portfolio is concentrated
in the trade sector and in the Baku area. With respect to the total number of clients,
AccessBank is with 63,432 business loan clients in July 2009 the leading bank in Azerbaijan
and has a total market share of 38% in the microfinance sector. The bank has been profitable
since 2004 and in 2008 the return on equity was at 44.4 percent. Despite the crisis, the
portfolio quality is very good with a portfolio at risk (PAR>30 days) of below 1 percent in
mid 2009.


3.2 The borrower groups
    While AccessBank also offers retail and staff loans, its focus and thus the focus of our
study is on business lending. Within the business portfolio, agro, micro and SME loans are
offered. The agro loan product was developed in 2007 to target farmers and the agricultural
sector by offering flexible disbursement and repayment schemes tied to the agricultural cycles
and the respective cash-flows of agro-businesses. The agro clients of AccessBank mainly
produce food crops and livestock for local markets so that the demand for their products
should be relatively unaffected by the crisis. Micro loans have amounts up to 10,000 USD
(from 2008 onwards up to 20,000 USD) and SME loans have amounts of more than 10,000
USD (20,000 USD) and up to 500,000 USD.
    With respect to micro and SME borrowers in general, most microenterprises are in the
business of subsistence goods and services for which demand does not decline much even in
times of crisis (Littlefield and Kneiding (2009)). In Azerbaijan, as in other emerging markets,
microenterprises are predominantly active in small-scale retail trade or transport. SME
businesses, in contrast, are involved in a broader variety of activities and rather tend to
provide non-essential goods and services (such furniture and household appliances or real
estate services), for which demand typically decreases in economic downturns. In contrast to
the large firms in the oil and gas industry, only very few MSMEs in Azerbaijan participate in
foreign trade activities (IFC (2009)). However, SME borrowers are more likely than micro
borrowers to have some of their income in foreign currency and are generally better able to
deal with foreign currency risk because of their higher income streams (AccessBank (2008)).
    The two products not only differ by their loan sizes (and the general characteristics of the
firms taking them out) but also with respect to the loan granting process. While a personal


                                                                                               9
visit of the loan officer and an analysis of the business finances as well as the debt servicing
capacity are part of the rigorous financial analysis of all business clients, the process is more
structured for SME loans including the consideration of longer time-periods of data and
financial projections. Therefore, the loan terms of SME loans capture a broader range of
individual risk factors, whereas loan terms for micro loans are more standardized due to their
small amounts. That the screening process for SME loans is more extensive becomes also
apparent from the fact that the time between loan applications and the approval decision is
longer for SME than for micro loans.
    Given the differences between the various loan types and the clients they target, we treat
agro, micro and SME loans separately throughout our analysis.


3.3 The dataset
    In total, the bank disbursed 251,211 loans to 151,533 clients since the start of its
operations. We exclude all observations with missing loan or firm characteristics as well as all
retail loans to private households and loans to bank staff to focus on the bank’s main client
groups. Table 1 provides a detailed overview of the bank’s business loan portfolio since its
start of operations. While Panel A shows the number of disbursed loans by year for our three
loan categories of agro, micro and SME loans and the number of rejected loans, Panel B
displays the respective total volumes of the loans disbursed. The majority of loans in our
sample are micro loans (75% in terms of numbers and 56% in terms of volume). However,
when considering total loan volumes, it becomes clear that SME loans make up a sizable part
of the bank’s business loan portfolio with a share of 35%. Agro loans, which were introduced
in 2007, seem to play an increasingly important role in the bank’s lending business. Rejection
rates were substantial in the beginning of the bank’s operations but have come down to less
than 6% in 2009. One explanation is that the bank deals with more and more repeat clients
over time so that it can assess their credit risk better due to reduced informational
asymmetries. At the same time, (potential) borrowers might have become acquainted to the
bank’s loan granting standards and have learnt to better self-assess whether their loan
application will be successful (see Kirschenmann (2011) for borrower learning in repeated
interactions with the same lender).


                                      [Insert Table 1 here]




                                                                                              10
     For each loan the dataset includes information on the loan amount and currency
requested by the borrower as stated in the loan application form as well as the granted loan
terms (amount, maturity and collateral) as stated in the loan contract. Furthermore, the data
not only contains information on the actual clients, but also on those who applied for a loan,
but were rejected. Table 2 provides definitions of all variables.


                                            [Insert Table 2 here]


     To examine relatively balanced time periods around the crisis shocks and to make sure
that our results are not driven by possible procedural changes during the early years of the
bank’s operations, we concentrate our empirical analysis on those loans disbursed from the
beginning of 2006 onwards until the end of our observation period in August 2009. 11 We also
focus on the branches that were established before the crisis so that we have observations for
both time periods (before and after the various crisis events) and our results are not biased
because of a change in the geographical distribution of customers. This final sample consists
of 168,483 loan applications from 88,370 firms.


3.4 The timeline of events and the impact of the crisis on credit availability
     Since we observe both loan applications and actually granted loans we are able to
establish the impact of the financial crisis on borrowers’ requests and the bank’s decision to
approve or reject the loan application and to assess whether the three subgroups of loans are
affected differently. Figure 1 displays the timeline of events related to the international
transmission of the financial crisis to AccessBank. In particular, two major crisis events that
may influence loan demand and supply decisions occurred in 2008.
     Firstly, AccessBank experienced unexpected, temporary delays in its refinancing pipeline
during the second and third quarters of 2008 so that it did not have as much funding as would
have been needed to meet total credit demand. These delays can be considered unexpected
because AccessBank’s financial performance was strong throughout our observation period.
More specifically, with respect to debt refinancing, the bank attracted funds from various
international financial institutions (not only the shareholders) and received two funds
denominated in Manat from private capital sources. AccessBank has been rated BB+ by Fitch
Ratings – the highest rating in Azerbaijan – which helped it to complete the first bond issue
on international capital markets by an Azeri issuer in February 2008. Also, the deposit
11
 We also run all of our regressions with the full sample of loans extended since the bank’s start of operations in
October 2002 and find our main conclusions to hold.

                                                                                                               11
portfolio was growing. Thus, the temporarily tight refinancing was not at all caused by a
change in the investors’ perception of the institution but solely by the turmoil in capital
markets which induced the cancellation of and delay in international refinancing. International
capital markets were unable to provide the necessary liquidity because, for instance, a planned
CDO could not be placed in the prevailing environment.
    Together with the strong portfolio growth, these unexpected funding difficulties forced
the bank to slow down its lending activities while the economy (as most other emerging
market economies) was still growing. At the same time, it successfully undertook efforts to
secure additional, initially not planned, financing and the funding situation relaxed from the
fourth quarter of 2008 onwards although the financial crisis had reached its peak with the
failure of Lehman Brothers. At first sight, this may seem counterintuitive but (commercial)
investors in microfinance have shown a “flight to quality� in times of crisis (Standard &
Poor's (2008)) corroborating the overall excellent performance of AccessBank during the
crisis. Summarizing, this first event can be considered as a pure funding or supply side shock
while there was no confounding shock to the demand side.
    Generally, one would expect this event to mainly influence the bank’s ability to meet
given loan requests. However, the tight refinancing may also mirror itself in the number of
loan applications because of the particular marketing strategy of AccessBank (and similar
MSME banks). All in all, it seems that MSMEs in Azerbaijan, and those that have never had a
loan before in particular, do not know much about the services banks offer and complain
about their terms and conditions although only knowing them by word of mouth (IFC (2009)).
To attract new borrowers, the bank therefore regularly organizes local community events to
market its products (AccessBank (2007)). During the time of tight refinancing, loan officers
were encouraged to take vacation which is equal to curtailing the active marketing of the
products and thus the number of loan applications may also decline due to the temporary
squeeze in the bank’s liquidity. Therefore, we expect the tight refinancing to have a negative
effect on all firms’ (agro, micro and SME) credit availability via the discouragement of
potential borrowers. We do not expect to find an effect on loan approval rates because the
economy was still largely unaffected by the crisis.


                                       [Insert Figure 1 here]


    Secondly, together with the sharp decline in oil prices, the economic crisis finally hit
Azerbaijan in the fourth quarter of 2008, exactly at the time when the funding situation of


                                                                                            12
AccessBank started to normalize. Some first spillovers of the economic crisis already
occurred during the summer of 2008 when oil prices and exports started to decline. However,
the sharp contraction of exports happened in the fourth quarter of 2008 and the first quarter of
2009 with exports to the neighboring CIS countries exhibiting a more sustained decline than
total exports (see Appendix 1 for details). At the same time, households’ income streams were
affected by the effects of this global crisis as remittances fell by around 20% in 2009
compared to 2008 (World Bank (2011)). Together with the beginning slump of the retail car
market and decreasing apartment sales reported in November 200812 this implies that the
demand for products, non-subsistence goods (and services) in particular, declined from the
fourth quarter of 2008 onwards. Finally, bank credit to households contracted by around 2%
in 2009 compared to a rapid expansion during the years before (Economist Intelligence Unit
(2010)). Turning to the bank’s marketing efforts during this time period, in contrast to the
period of tight refinancing, the availability of business loans was explicitly and actively
marketed in 2009 to show AccessBank’s ability and willingness to be a reliable partner also
during the crisis (AccessBank (2009)).
     Taking all this together, the effect of the real economy shock on the number of loan
applications is ambiguous. On the one hand, loan applications could increase at AccessBank
because of the liquidity problems and the reduced lending at other banks. On the other hand,
loan applications could decrease because business borrowers saw their sales decline and thus,
for instance, postponed investment. However, in contrast to the first event, we expect
approval rates to decrease because of the worsening economic situation. Screening of
borrowers should have become more important to identify the less risky as borrower quality
in general started to deteriorate. We expect to find a differential effect for agro, micro and
SME borrowers. SME loans are likely to be affected most because SME firms often produce
non-subsistence goods or services or may even be suppliers to the oil industry and are
therefore more likely to be hit by the economic slowdown and to see their credit quality
deteriorate.
     Our empirical analysis will focus on these two distinct crisis events with the pre-crisis
period comprising the years 2006 and 2007.


3.5 Determinants of credit approval




12
  See Economist Intelligence Unit (2009a) and
http://www.eurasianet.org/departments/insightb/articles/eav120108a.shtml.

                                                                                             13
     We analyze the effects of the 2007-2009 crisis on the availability of credit for our sample
of agro, micro and SME loans with a linear probability model (LPM)13 in which the
dependent variable Pr(Approved)i, k, t is the probability of firm i to receive loan k in period t:


       Pr(Approved)i,k,t = αi + β1Ck,t + β2Lk,t + β3Fi,t + εi,k,t                                            (1)


     Ck,t is a dummy variable that indicates whether loan k was disbursed during one of the
crisis events (i.e. the funding or the real economy shock) while Lk,t and Fi,t are vectors of loan
and firm characteristics including 6 bank branch dummies that control for the location of loan
origination.


Crisis indicators and loan characteristics
     In line with the timeline of the crisis depicted in Figure 1, we include two dummy
variables to compare each of the two distinct crisis periods of interest with the non-crisis
years. Funding shock equals one during the months of tight refinancing between April and
August 2008 and zero in the non-crisis years 2006 and 2007.14 Real economy shock, in
contrast, equals one for those loans disbursed from October 2008 onwards and zero for the
loans disbursed during the non-crisis years 2006 and 2007. We account for the structure of the
bank’s loan portfolio by introducing the dummy variables Agro, Micro and SME which are
one if the loan is an agro, micro or SME loan, respectively.
     We include Repeat loan as a relationship indicator since close bank-borrower
relationships have been found to increase credit availability (e.g. Petersen and Rajan
(1994)).15 It is a dummy variable indicating whether a loan is a firm’s later vs. first loan. We
expect the bank to be able to gather valuable private information when interacting repeatedly
with the same borrower, which, in turn, may benefit the borrower (see Allen and Gale (1999),
Boot (2000) and Ongena and Smith (2000)). Additionally, if a firm requests a very high loan
amount, the bank may be more likely to reject the loan application or to grant a considerably
lower loan amount. Therefore, we include the variable Requested amount which is the

13
   We re-estimate all our regressions using probit and logit models. Since coefficients yield the same signs, we
restrict the presentation of results to the LPM models due to the difficulty in deriving marginal effects in non-
linear regressions with interaction terms (see Ai and Norton (2003)).
14
   Defining the pre-crisis period until August 2007 does not change our main conclusions but does not allow us
to study the impact of the crisis on agro loans as these were only introduced in August 2007. The first quarter of
2008 is left out from the refinancing shock analysis to make sure that periods are clearly marked-off from each
other. Yet, our results are not sensitive to assigning Q1 2008 to the non-crisis period.
15
   We repeat all estimations with two other relationship measures, the length of the bank-borrower relationship
and the number of times a borrower has taken out a loan before the current loan, and find that results remain
qualitatively unchanged.

                                                                                                              14
requested loan amount in USD. The bank’s decision to grant a loan and its willingness to
grant the requested amount should critically depend on the perceived risk of the loan. One
means to make a loan safer is to pledge collateral. We include the variable Collateral which is
the value of the pledged collateral as a share of the requested amount (in %). The dummy
variable Manat requested indicates whether a loan is requested in the local currency Manat
vs. in foreign currency (USD).


Firm characteristics
     We include several firm characteristics that may influence the bank’s decision to grant a
loan. Since most of the firms in the sample are run by their owners, owner characteristics are
of particular importance. First of all, we include the borrower’s Age (in years) at the time of
the first loan with the bank. Older firm owners may be more experienced and therefore less
risky borrowers, which would mean a positive relation between borrower’s age and credit
availability. In microfinance, often the argument is made that women are better borrowers
because they are more reliable in repaying their loans (e.g. Armendariz de Aghion and
Murdoch (2005)). To capture a possible gender effect, we include the dummy variable Male
which is one if the borrower is male and zero if she is female. A similar reasoning might be
true for a Married (dummy variable that is one if the borrower is married and zero otherwise)
borrower who has responsibility towards a family so that we expect to find a positive relation
with credit availability.


3.6 Summary statistics
     Table 3 presents summary statistics of our loan and firm variables for the approved loans.
Panel A provides statistics for the full sample, whereas Panel B displays sample means by
loan category for the periods before and during the Real economy shock.
     Panel A shows that most of the loan applications in our sample are successful (Approved)
which may be another indication of informal discouragement of borrowers before they
actually fill in a loan application. The statistics confirm that the majority of approved loans in
our sample are micro loans with an average Requested amount of 3,986 USD. Interestingly,
loans are regularly collateralized at clearly more than 100% (Collateral). Considering firm
characteristics, the majority of borrowers are Male (83%) and Married (72%).


                                      [Insert Table 3 here]



                                                                                               15
    Panel B of Table 3 presents statistics for the three loan categories of agro, micro and
SME loans. To assess changes in loan terms and the borrower pool along the crisis, it
compares loan and firm characteristics of approved loans for the time before and during the
Real economy shock. The statistics confirm that the availability of agro loans in terms of
approval rates was not negatively affected by the Real economy event. For micro loans, the
impact of the crisis on the credit availability is relatively small as well. For SME loans, in
contrast, the probability that a loan application is Approved decreases by 6 percentage points
after the economic crisis hit. The average Requested amount for all loan types increases
regardless of the economic downturn. This is similar to Berg and Schrader (2011) who find
increased demand of micro borrowers after unexpected external shocks such as volcanic
eruptions. Collateral, in contrast, decreases for all loan types after the economic crisis hit
which may be due to the fact that real estate prices dropped considerably. However, on
average loans are still collateralized by much more than 100% during the economic downturn.
    Establishing good bank-borrower relationships seems to be especially important for
micro and SME clients as the share of Repeat loans increases for both loan groups during the
economic crisis. This confirms the graphical analysis in Figure 2 that existing bank-borrower
relationships help to enhance credit availability in times of crises. Interestingly, the economic
crisis influences the denomination of agro and micro loans considerably. After the onset of
the economic downturn 90% of agro loans and 69% of micro loans are requested in local
currency (Manat) compared to 49% and 41% before. This finding may at least partly be
explained by the fear of a depreciation of the local currency towards the USD after the
currencies of neighboring countries such as Russia and Kazakhstan plunged considerably due
to the crisis, even though the Manat did in fact only depreciate little. The statistics on firm
characteristics imply that the pool of borrowers is very similar in this respect before and
during the economic crisis.


4 Results
4.1 Graphical analysis of loan applications and approvals
    Figures 2 and 3 display the number of loan applications (168,483 in total) and loan
approvals (158,481 in total) and the respective approval rates during the period 2006 to 2009
for new vs. repeat borrowers in our three subsamples respectively.


                                 [Insert Figures 2 and 3 here]



                                                                                              16
     Figure 2 shows that loan applications and approvals develop differently for the three
subgroups. Agro loan applications drop considerably between April and June 2008, but
interestingly, repeat borrowers are much less discouraged.16 After this short period of
retrenchment loan applications again steadily increase for both new and repeat borrowers
(Figure 2a). This is a first indication that the agricultural sector remained mostly unaffected
by the economic crisis. Nevertheless, the bank’s tight refinancing clearly affects agro lending
because clients were obviously deterred from applying for loans. Loan approval rates
(approved loans as a share of applied loans) decrease somewhat, for the new borrowers in
particular, during the period of the tight refinancing but remain at high levels of more than
94% afterwards (Figure 3a).
     For micro loans, we similarly observe a considerable decrease in loan applications in the
second quarter of 2008 for new and repeat loans (Figure 2b). The approval rate analysis
(Figure 3b) shows that new micro loans are on average about 4 percentage points less likely to
be approved than repeat micro loans and that approval rates are more volatile for new than for
repeat micro borrowers. However, while the discouragement effect seems to be smaller for
repeat borrowers, their approval rates decrease slightly after the economic crisis hit. These
findings suggest that both the bank’s liquidity squeeze as well as the economic downturn
affected micro borrowers’ credit availability
     Figure 2c reveals that the demand for new and repeat SME loans is more volatile;
however there is also a clear decline in the number of loan applications after March 2008.
Considering the approval rates of SME loans (Figure 3c), we observe a negative trend after
the start of the economic crisis. Approval rates for SME loans decrease much more than for
agro and micro loans from around 90% to 50% for new SME loans and from above 95% to
75% for repeat SME loans. This indicates that the comparatively bigger firms in our sample
are more seriously hit by the crisis. There are several potential reasons for these findings:
First, during the tight refinancing period it may have been easier and cheaper for the bank to
“save� a certain amount of liquidity by discouraging as well as denying some SME loans in
contrast to an even larger number of micro loans. Second, the larger firms may have had to
cope with a larger and more persistent decrease in demand for their products once the
economic crisis hit implying that lending to the more heterogeneous SMEs became more



16
  Partly the drop may be due to seasonal effects as agro clients’ income streams usually peak in spring and
summer. However, if the decrease was only driven by cyclical patterns in agriculture, effects should have been
similar for new and repeat agro borrowers (as is the case exactly one year later during the second quarter of
2009). Apart from that, micro and SME loans show a sharp decrease in loan applications starting at exactly the
same point in time.

                                                                                                           17
risky than lending to agro and micro businesses.17 Third, SME clients are more likely to
finance fixed assets which are first to be postponed in times of crises while micro clients often
take out working capital loans. And fourth, with a decrease in housing prices in Azerbaijan,
borrowing to finance real estate became less attractive.
     The graphical analysis in Figure 2 suggests that both external events have an impact on
the lending operations of AccessBank but that these effects vary for the three different types
of loans. While credit availability for agro borrowers decreases for a very short period during
the refinancing delays, credit availability for micro borrowers is tightened moderately and for
SME borrowers most persistently.18 Apart from that, Figure 2 establishes the different
channels through which the funding and the real economy shock influence credit availability.
The refinancing delays seem to mainly work via the discouragement channel (i.e. less active
marketing, loan officers on holidays), whereas the real economy effect works via the loan
approval channel as well.
      Moreover, the analysis shows that previous bank relationships help borrowers to
mitigate negative crisis effects which is in line with the findings of, for instance, Berg and
Schrader (2011) and Puri, Rocholl and Steffen (2010). This relationship effect works mainly
through the discouragement channel, i.e. repeat borrowers are less likely to be discouraged. If
they apply for a loan, their approval likelihood is generally higher than that for new
borrowers, but it also decreases during the economic downturn.
     As the structure of our dataset also allows us to measure credit availability on an
individual level, the following analysis will be concerned with the impact of the crisis on loan
approval rates.


4.2 Funding shock and the likelihood that a loan application is approved
     We start our analysis by studying the effect of the unexpected refinancing delays on the
likelihood that a loan application is approved by the bank. Table 4 reports results from LPM
regressions for the full sample and our three subsamples based on our three loan categories.
Standard errors are reported in parentheses and are adjusted for clustering at the borrower
level. All coefficients are multiplied by 100 so that they can directly be read as percentages.

17
   Interestingly, Figure 2 also shows that loan applications of new borrowers increase after the onset of the
economic slowdown, while loan applications from repeat borrowers decrease. This is in line with our reasoning
that loan applications may actually increase during the economic downturn because of the liquidity problems of
other Azeri banks. However, the decline in approval likelihoods for both new and repeat SME clients indicates
the increased riskiness of SME borrowers in general. This also indicates that intense screening is the adequate
mechanism to deal with a shock that is a pure real economy shock for AccessBank but also a liquidity shock for
the other banks in the country.
18
   Analyzing total loan volumes and volume constraints also reveal that SME borrowers are most affected by the
crisis.

                                                                                                            18
The results confirm the evidence from the graphical analysis and show that the refinancing
shock mainly impacted on firms’ credit availability via the discouragement channel. Those
loans that were actually applied for have, if anything, a higher likelihood to be approved
during the months of tight refinancing compared to the non-crisis years of 2006 and 2007.
    The estimates in Column (1) show, for the full sample, that the approval likelihood for
loan applications of SME borrowers (the base category) made during the Funding shock is 2.9
percentage points higher than their approval likelihood during the non-crisis period. Agro and
Micro loan applications are significantly more likely to be successful than SME loan
applications during the non-crisis years. On average, agro (micro) loan applications have an
8.4 percentage point (5.0 percentage point) higher probability to be finally granted compared
to SME loan applications in 2006 and 2007. To study how the relation between our loan
groups and the probability of loan approval is influenced by the funding shock, we study the
interaction terms Agro*Funding shock and Micro*Funding shock. Both interaction terms are
negative but only marginally significant and smaller than the main effects. This means that,
during the time of tight refinancing, the difference in the probability to be granted a loan
decreases somewhat when comparing agro and micro to SME loans but the former still have a
higher approval likelihood.
    Columns (2) to (4) report estimates of LPM regressions for agro, micro and SME loans
respectively to further disentangle differences in the factors determining loan approval rates
for these three categories. Now, the main effect of our Funding shock indicator displays the
impact of the unexpected refinancing delays on new loans’ approval likelihood showing that
the Funding shock has no negative impact at all on loan approvals for those new loans that
were actually applied for. The significantly positive main effect of Repeat loan implies that
being a repeat borrower is crucial for increasing credit availability for all three loan groups in
the non-crisis period with the effect being most pronounced for SME loan. Since SME loans
involve considerably larger loan amounts, the information gathered from previous interactions
with borrowers may be most valuable in this segment.
    The interaction term Repeat loan*Funding shock is insignificant for agro and
significantly negative but very small for micro loans. This indicates that, whether in good or
bad times, having a previous relationship with the bank is similarly beneficial in increasing
the approval likelihood after applying for a loan. For SME loans, the interaction term Repeat
loan*Funding shock is significantly negative but economically smaller than the main effect of
Repeat loan. This means that repeat compared to new SME borrowers still have a higher
approval likelihood during the time of tight refinancing but the difference in approval


                                                                                               19
likelihoods becomes much smaller. This is in line with the interpretation of the graphical
findings that the relationship effect mainly works via less discouragement for the repeat
borrowers during the time of tight refinancing.
     The denomination of the loan influences credit availability adversely for the three loan
groups. While agro and micro loans are more likely to be approved when they are requested
in Manat, SME loans requested in Manat are less likely to be granted. One explanation for
this result lies in the bank’s funding structure. Since it receives most of its funding in USD, it
has an incentive to lend on in foreign currency to prevent currency mismatches on its own
balance sheet. At the same time, the bank seems to channel its funds according to the
borrowers’ abilities to deal with foreign currency risks since SMEs are likely to be more
capable to handle USD loans or even earn (some of) their income in USD (Brown,
Kirschenmann and Ongena (2011) provide similar evidence for micro and SME lending in
Bulgaria).
     Since the missing refinancing funds during the second and third quarters of 2008 were to
be denominated in USD which should have impacted on the bank’s preferred lending
currency we study the interaction term Manat*Funding shock for the three loan types in
unreported regressions (results are available from the authors on request). Indeed, this
analysis reveals that SME loans are less likely to be approved when requested in Manat only
during the non-crisis period but actually more likely to be approved when requested in Manat
during the time of tight refinancing. There is no such effect found for SME loans when
studying the interaction effect Manat*Real economy shock.
     Turning to the impact of the firm characteristics, we find that only the variable Married
has a uniform, significantly positive impact on loan approval rates for all three loan
categories. Of all our firm characteristics it seems to provide the strongest signal to the bank.
The borrower’s gender works in the opposite direction for SME and agro vs. micro loans. For
SME and agro loans, being Male decreases the probability to receive a loan. Yet, male
borrowers who request micro loans have a slightly higher probability to succeed in applying
for a loan in comparison to female borrowers. Finally, a borrower’s Age at first loan
disbursement plays a minor role in determining credit availability.19


4.3 Real economy shock and the likelihood that a loan application is approved
     Table 5 re-runs our regressions from Table 4 but concentrates on the effect of the real
economy shock on the bank’s decision to approve a loan application. As before, standard
19
   Including Age squared as an additional variable, we find Age to be positive and Age squared negative, but
effects are either insignificant or very small.

                                                                                                         20
errors are reported in parentheses and are adjusted for clustering at the borrower level. All
coefficients are multiplied by 100 so that they can directly be read as percentages. The results
confirm the main hypothesis: the real economy shock has, in contrast to the funding shock,
some negative impact on credit approval rates, but the effect varies significantly between
agro, micro and SME loans.


                                                [Insert Table 5 here]


        More specifically, in Column (1) the variable Real economy shock shows that SME loan
applications made from October 2008 onwards compared to the non-crisis years have, on
average, a 6.1 percentage point lower chance to be approved. The results on the differential
impact of the real economy shock on loan approval in the three loan groups differ
considerably from the funding shock results with both main effects and both interaction terms
being significantly positive. This means that the probability to be granted a loan is not only
higher for Agro and Micro loans compared to SME loans in the non-crisis years but that this
effect is even more pronounced after the onset of the economic downturn. For instance,
during the non-crisis years Agro loans are 6.8 percentage points more likely to be approved
than SME loans. During the economic crisis, this difference in probabilities increases by
another 6.6 percentage points. The effects are qualitatively the same for Micro loans, yet they
are economically smaller.
        The separate regressions for agro, micro and SME loans respectively in columns (2) to
(4) of Table 5 include the interaction term Repeat loan*Real economy shock to assess whether
previous bank relationships are especially beneficial during the economic crisis. The main
effect of our Real economy shock indicator displays the impact of the economic downturn on
new loans’ approval likelihood. It is positive for agro and micro loans, however for the latter
it is negligible in economic terms. For new SME borrowers, in contrast, the probability to
receive a loan decreases during the economic crisis by 7 percentage points. As in the case of
the tight refinancing period, we find that being a repeat borrower increases the likelihood of
loan approval similarly in good as in bad times.20 The results for the other loan and firm
characteristics are qualitatively very similar to Table 4.
        Summarizing, we find that the impact of the economic crisis on credit availability
strongly depends on firm size and on the industry a firm operates in. These results relate to the
specific structure of Azerbaijan’s economy with its larger firms being more dependent on oil

20
     The interaction effect is significantly negative for micro loans, yet it is economically not very large.

                                                                                                                21
price fluctuations and facing more variable demand for their products. Micro firms, on the
contrary, rather produce subsistence goods or deliver essential services and agro borrowers
mostly grow fruit and vegetables or raise sheep for the local market so that both are
considerably less affected by economic fluctuations. These varying levels of affectedness
seem to influence the bank’s risk assessment of firms and to finally translate into accordingly
varying levels of credit availability.


4.4 Geographical variation in the crisis shock
        To confirm that the two shocks are distinctly different, we use the location of the branch
that granted the loan, which is also a proxy for the location of the firm, to study exogenous
variation in the severity of the crisis shock in the following analysis. Given that Azerbaijan’s
economy is weakly diversified and highly dependent on the oil and gas sector, we distinguish
between Petroleum locations, i.e. locations in which the oil and gas industry is most
important according to the State Statistical Office of the Republic of Azerbaijan 21 and the
location of the major oil and gas fields, and the rest of the country. These petroleum locations
are the capital city of Baku and the close-by next biggest city Sumqayit. We then analyze the
interaction effects Funding shock*Petroleum location and Real economy shock*Petroleum
location. We expect the latter interaction term to be significantly negative as the petroleum
locations should be affected more severely by the economic downturn. In contrast, the
funding shock led to short funding which should have restricted lending in all branches over
the country. Therefore, we expect the interaction term Funding shock*Petroleum location to
be statistically insignificant.
        Table 6 displays results for LPM regressions on the bank’s decision whether to approve a
loan application or not for the full sample and the three subsamples based on our loan type
categories. Columns (1-4) present the results for the effect of the funding shock, while
columns (5-8) contain the respective results for the real economy shock.


                                              [Insert Table 6 here]


        As expected, the interaction term Funding shock*Petroleum location is not statistically
significant in the full sample as well as for agro and micro loans. Together with the
significantly positive main effect Funding shock this indicates that for loans made in non-
petroleum locations approval rates for those borrowers who actually apply for a loan are

21
     See www.azstat.org/statinfo/industry/en/index.shtml#.

                                                                                               22
higher during the time of tight refinancing in comparison to the non-crisis years, while this
effect is not significantly different in petroleum locations. For SME loans, in contrast, the
interaction effect displayed in column (4) is significantly negative but economically smaller
than the main effect of Funding shock. This means that SME loans made in petroleum
compared to non-petroleum locations still have a higher approval likelihood during the time
of tight refinancing but the difference in approval likelihoods becomes much smaller.22
     With respect to the real economy shock, we additionally deepen the analysis by dividing
its effect into a short-term (Real economy shock 2008) and a long-term (Real economy shock
2009) component capturing the respective months in 2008 or 2009 after the onset of the
economic downturn.23 For the full sample in column (5), the very small positive main terms
of Real economy shock 2008 and Real economy shock 2009 together with the larger and
significantly negative interaction effects show that loans made in petroleum locations have a
significantly lower approval likelihood after the start of the economic slowdown than loans
made in non-petroleum locations. The latter finding also holds when studying the three loan
categories separately in columns (6-8). For SME loans in column (8) the interaction terms are
not statistically significant but economically relevant. We find that in the early months of the
economic slowdown the approval likelihood of SME loans in petroleum locations also
decreases, while, in contrast to agro and micro loans, this decrease is reversed again in the
longer-term. Interestingly, we also find that SME borrowers seem to be hit by the crisis in the
whole country, whereas agro and micro borrowers are only affected in petroleum locations as
for the latter two groups the main effects Real economy shock 2008 and Real economy shock
2009 are positive. For SME borrowers, in contrast, the approval likelihood for loans granted
in non-petroleum locations also decreases significantly and considerably with some delay in
2009 compared to the non-crisis years.
     These results confirm our hypothesis that the real economy shock is distinctly different
from the funding shock. The real economy shock affects credit availability negatively
especially in those locations that can be expected to be hit hardest, i.e. the petroleum
locations. The funding shock, on the contrary, has no negative and no differential (except for
SME borrowers) effect for petroleum- vs. non-petroleum locations. This confirms our
previous finding that the refinancing delays are a pure supply shock impacting on credit
availability mainly through the discouragement channel.

22
   This result is mainly driven by the fact that for the 10% of SME loans that are actually requested in non-
petroleum locations the approval likelihood increases considerably during the time of tight refinancing, while
this increase is only very moderate in the petroleum locations where 90% of SME loans are applied for.
23
   Our conclusions are unaltered if we refrain from this additional time split, but the differential short-term vs.
longer-term effect for SME loans would remain hidden in this analysis.

                                                                                                               23
5 Conclusions
    This paper studies the impact of the financial crisis on the credit availability for micro,
small and medium enterprises in Azerbaijan using a unique dataset of 168,483 loans made
between January 2006 and August 2009 by AccessBank Azerbaijan. The timeline of events in
Azerbaijan allows us to study a pure supply side shock to the bank’s refinancing in the second
and third quarters of 2008, while firms had not yet been affected by the economic crisis.
Then, we can contrast its impact on credit availability for firms of different sizes and
industries from the impact of the economic crisis shock that reached Azerbaijan in the fourth
quarter of 2008. On top of this, the structure of the dataset allows us to analyze credit
availability on an aggregate level as well as on an individual loan level, thereby identifying
two major channels of credit constraints.
    By analyzing aggregate numbers of loan applications and approvals, we derive the first
channel of credit constraints. We observe that its tight refinancing in the second and third
quarters of 2008 seems to lead the bank to discourage (potential) borrowers from applying for
new or additional loans. The bank’s temporary liquidity squeeze therefore resulted in a
slowdown of its credit portfolio so that not all the demand could be met during this limited
period of time. At the individual loan level, for the funding shock we find that approval rates
for those (potential) borrowers who actually apply for a loan are not negatively affected in any
borrower group. For the real economy shock, by contrast, our findings suggest that credit
availability for agro loan borrowers is hardly affected by this event while micro loans face a
moderate and SME loans a considerable reduction in approval rates. Apart from that, our
results show that bank relationships help increase credit availability in good as in bad times.
    To gain deeper insights into the differential effects of the two crisis shocks on loan
approval, we study an exogenous geographic variation in the severity of the crisis shock by
distinguishing between locations in which the oil and gas industry is most important and the
rest of the country. These results confirm that the two shocks differently impact on credit
availability. The real economy shock affects credit availability negatively especially in those
locations that can be expected to be hit hardest, i.e. the petroleum locations. Our results for
the funding shock, on the contrary are in line with our previous finding that the refinancing
delays are a pure supply shock impacting on credit availability mainly through the
discouragement channel.
    In contrast to anecdotal evidence from some more advanced and competitive markets in
Eastern Europe (e.g. ProCredit Holding (2009)) which suggests that banks worried especially


                                                                                                  24
about their lending to micro clients and limited their exposure in that segment because many
of these clients carried high levels of (consumer) debt, we find that credit availability is
mostly affected for SME customers. However, one caveat has to be made concerning the
interpretation of our results. To comprehensively assess the effects of the crisis on credit
availability for the various types of enterprises one would have to more explicitly take into
account how many borrowers have been deterred from applying for a loan by their loan
officers. While we can provide some evidence on this aspect in our aggregate analysis, it is
possible that the active marketing strategy which the bank uses to attract new customers is
more intensively employed in the segment of micro loans compared to SME loans. This
would imply that a decrease in these marketing activities would have a larger impact on the
micro than on the SME loan portfolio and we would underestimate the negative effect of the
crisis on micro credit availability. This opens up room for future empirical research to
broaden the evidence on the effects of the financial crisis on credit availability.
    Our results have implications for policy makers and development practitioners aiming at
sustainably fostering credit access for micro, small and medium businesses in developing and
transition economies. First, supporting MSME banks in building up diversified credit
portfolios that include various loan categories with respect to size and industry may increase
these banks’ stability in times of a global financial and economic crisis as the 2007-2009 one.
However, further research on how different banks’ portfolio quality is affected by such a
crisis would be needed to shed more light on this aspect. Second, broadening MSME banks’
refinancing basis to achieve greater resilience against external shocks remains an important
topic. Recent attempts to create adequate (long-term) refinancing instruments in local
currency therefore seem to be a crucial step to help MSME banks to overcome refinancing
problems.




                                                                                            25
References

AccessBank (2007). "Accessbank Annual Report 2007". Baku, Azerbaijan.
AccessBank (2008). "Accessbank Annual Report 2008". Baku, Azerbaijan.
AccessBank (2009). "Accessbank Annual Report 2009". Baku, Azerbaijan.
Ai, C. and E. C. Norton (2003). "Interaction Terms in Logit and Probit Models". Economics
        Letters 80, 123-129.
Allen, F. and D. Gale (1999). "Innovations in Financial Services, Relationships, and Risk
        Sharing". Management Science 45, 1239-1253.
Armendariz de Aghion, B. and J. Murdoch (2005). "The Economics of Microfinance".
        Cambridge, MA: MIT Press.
Bankakademie (2000). "Marketing for Microfinance Depositories". GTZ, Eschborn,
        Germany.
Bekaert, G., C. R. Harvey and C. Lundblad (2005). "Did Financial Liberalization Spur
        Economic Growth?". Journal of Financial Economics 77, 3-55.
Berg, G. and J. Schrader (2011). "Access to Credit, Natural Disasters, and Relationship
        Lending". Working Paper.
Boot, A. (2000). "Relationship Banking: What Do We Know?". Journal of Financial
        Intermediation 9, 7-25.
Brown, M., K. Kirschenmann and S. Ongena (2011). "Securitization, Bank Funding and
        Loan Terms: Evidence from Foreign Currency Lending". EBC Discussion Paper No.
        2009-21.
Brown, M., S. Ongena, A. Popov and P. Yesin (2011). "Who Needs Credit and Who Gets
        Credit in Eastern Europe?". Economic Policy 26, 93-130.
Brunnermeier, M. (2009). "Deciphering the Liquidity and Credit Crunch 2007-2008". Journal
        of Economic Perspectives 23, 77-100.
Central Bank of the Republic of Azerbaijan (2008). "Annual Report 2008". Baku, Azerbaijan.
Central Bank of the Republic of Azerbaijan (2009). "Statistical Bulletin March 2009". Baku,
        Azerbaijan.
Christen, R. (2001). "Commercialization and Mission Drift - the Transformation of
        Microfinance in Latin America". CGAP Occasional Paper No. 5.
De Haas, R., Y. Korniyenko, E. Loukoianova and A. Pivovarsky (2012). "Foreign Banks and
        the Vienna Initiative: Turning Sinners into Saints?". MoFiR Working Paper No. 62.
De Haas, R. and N. Van Horen (2010). "The Crisis as a Wake-up Call. Do Banks Tighten
        Screening and Monitoring During a Financial Crisis?". DNB Working Paper No. 255.
De Haas, R. and N. Van Horen (2011). "Running for the Exit: International Bank Lending
        During a Financial Crisis". DNB Working Paper No. 279.
De Haas, R. and I. Van Lelyveld (2011). "Multinational Banks and the Global Financial
        Crisis: Weathering the Perfect Storm?". DNB Working Paper No. 322.
Di Bella, C. G. (2011). "The Impact of the Global Financial Crisis on Microfinance and
        Policy Implications". IMF Working Paper No. 11/175.
Dooley, M. and M. Hutchison (2009). "Transmission of the U.S. Subprime Crisis to
        Emerging Markets: Evidence on the Decoupling-Recoupling Hypothesis". Journal of
        International Money and Finance 28, 1331-1349.
Economist Intelligence Unit (2009a). "Azerbaijan: Financial Services Report, January 19th
        2009". London, UK.
Economist Intelligence Unit (2009b). "Country Report Azerbaijan July 2009". London, UK.
Economist Intelligence Unit (2010). "Azerbaijan: Financial Services Report, January 18th
        2010". London, UK.
Galema, R., R. Lensink and L. Spierdijk (2011). "International Diversification and
        Microfinance". Journal of International Money and Finance 30, 507-515.

                                                                                        26
Giannetti, M. and S. Ongena (2009). "Financial Integration and Firm Performance: Evidence
         from Foreign Bank Entry in Emerging Markets". Review of Finance 13, 181-223.
Guiso, L., P. Sapienza and L. Zingales (2004). "Does Local Financial Development Matter?".
         Quarterly Journal of Economics 119, 929-970.
Huang, R. (2010). "How Committed Are Bank Lines of Credit? Evidence from the Subprime
         Mortgage Crisis". Federal Reserve Bank of Philadelphia Working Paper No. 10-25.
Hübner, G. and M. Jainzik (2009). "Splendid Isolation? Azerbaijan's Economy between
         Crisis Resistance and Debased Performance". Caucasus Analytical Digest. Bremen,
         Germany and Zurich, Switzerland.
IFC (2009). "Study of Small and Medium Enterprises in Azerbaijan". Baku, Azerbaijan.
IMF (2008). "Azerbaijan - Aide Mémoire for the 2008 IMF Staff Visit Discussions".
         http://www.imf.org/external/np/ms/2008/121608b.htm.
Ivashina, V. and D. Scharfstein (2010). "Bank Lending During the Financial Crisis of 2008".
         Journal of Financial Economics 97, 319-338.
Jimenez, G., S. Ongena, J.-L. Peydro and J. Saurina (2012). "Credit Supply Versus Demand:
         Bank and Firm Balance-Sheet Channels in Good and Crisis Times". EBC Discussion
         Paper No. 2012-003.
Khwaja, A. I. and A. R. Mian (2008). "Tracing the Impact of Bank Liquidity Shocks:
         Evidence from an Emerging Market". American Economic Review 98, 1413-1442.
Kirschenmann, K. (2011). "Credit Rationing in Small Business Bank Relationships". Working
         Paper.
Littlefield, E. and C. Kneiding (2009). "The Global Financial Crisis and Its Impact on
         Microfinance". CGAP Focus Note No. 52, February 2009.
Littlefield, E. and R. Rosenberg (2004). "Microfinance and the Poor - Breaking Down Walls
         between Microfinance and Formal Finance". Finance and Development 41.
Ongena, S., J.-L. Peydro and N. Van Horen (2012). "Shocks Abroad, Pain at Home? Bank-
         Firm Level Evidence on Financial Contagion During the Recent Financial Crisis".
         Working Paper.
Ongena, S. and D. C. Smith. (2000). "Bank Relationships: A Review". In Harker, P. and S.
         A. Zenios (Eds.), "The Performance of Financial Institutions: Efficiency, Innovation,
         Regulation" (pp. 221-258). London: Cambridge University Press.
Paravisini, D. (2008). "Local Bank Financial Constraints and Firm Access to External
         Finance". Journal of Finance 63, 2161-2193.
Patten, R., J. Rosengard and D. Johnston (2001). "Microfinance Success Amidst
         Macroeconomic Failure: The Experience of Bank Rakyat Indonesia During the East
         Asian Financial Crisis". World Development 29, 1057-1069.
Petersen, M. A. and R. G. Rajan (1994). "The Benefits of Lending Relationships: Evidence
         from Small Business Data". Journal of Finance 49, 3-37.
Popov, A. and G. F. Udell (2011). "Cross-Border Banking, Credit Access, and the Financial
         Crisis". Journal of International Economics, forthcoming.
ProCredit Holding (2009). "Annual Report 2009". Frankfurt a.M., Germany.
Puri, M., J. Rocholl and S. Steffen (2010). "Global Retail Lending in the Aftermath of the US
         Financial Crisis: Distinguishing between Supply and Demand Effects". Journal of
         Financial Economics 100, 556-578.
Rajan, R. G. (1992). "Insiders and Outsiders: The Choice between Informed and Arm's-
         Length Debt". Journal of Finance 47, 1367-1400.
Standard & Poor's (2008). "Microfinance Institutions: Changing Strategies for Changing
         Times". New York, USA.
Transparency        International     (2008).   "Corruption    Perceptions    Index    2008".
         www.transparency.org/news_room/in_focus/2008/cpi2008/cpi_2008_table.


                                                                                           27
Wenner, M. D., S. Navajas, C. Trivelli and A. Tarazona (2007). "Managing Credit Risk in
      Rural Financial Institutions: What Seems to Work". Enterprise Development and
      Microfinance 18, 158-174.
World Bank (2011). "Migration and Remittances Factbook 2011". Washington, D.C., USA.




                                                                                    28
                           Table 1. Lending by year and loan type

This table reports statistics on the bank’s loan portfolio for the full sample and the following
subsamples for the whole period since the bank’s foundation in October 2002: Agro: Loans intended
for agricultural investments. Micro: Loans with loan amounts up to 10,000 USD (from 2008: also with
amounts up to 20,000 USD). SME: Loans with loan amounts of more than 10,000 USD and up to
500,000 USD. Rejected: Loan applications that were turned down by the bank.

                      Panel A. Number of loans disbursed and rejected


                       Rejected                      Disbursed                                Total
                                           Agro             Micro            SME
2002                         81                               104                              185
2003                        502                             2,891               66           3,459
2004                        424                             4,008              175           4,607
2005                        518                             7,224              436           8,178
2006                      1,362                            16,491              813          18,666
2007                      2,408            2,163           32,531            1,328          38,430
2008                      3,088           10,105           41,246            1,633          56,072
2009                      3,144           15,643           35,529            1,008          55,315
Total                    11,527           27,902         140,024             5,459         184,912




                      Panel B. Total volumes of loans disbursed (USD)


                                            Agro            Micro            SME              Total
2002                                                      113,077                           113,077
2003                                                    5,691,747        1,269,945        6,961,692
2004                                                   46,077,936        4,673,496       50,751,432
2005                                                   62,583,051       12,365,058       74,948,108
2006                                                   32,118,408       25,377,105       57,495,513
2007                                   3,103,264       58,840,534       44,415,355      106,359,153
2008                                  18,285,522      101,915,005       68,532,216      188,732,742
2009                                  27,117,810       90,049,888       40,972,025      158,139,722
Total                                 48,506,596      397,389,645      197,605,200      643,501,441




                                                                                                29
                                 Table 2. Variable definitions



Variable               Definition

Dependent variable
Approved               Borrower's loan request was approved (1=yes, 0=no)

Loan characteristics
Real economy shock     Loan was granted after the onset of the economic downturn (10/2008) vs. loan was
                       granted in 2006 or 2007 (1=yes, 0=no)
Funding shock          Loan was granted during the time of tight refinancing between 04/2008 and 08/2008
                       vs. loan was granted in 2006 or 2007 (1=yes; 0=no)
Agro                   Loan is an agro loan (1=yes, 0=no)
Micro                  Loan is a micro loan with an amount up to 10,000 USD (20,000 USD from 2008
                       onwards) (1=yes, 0=no)
SME                    Loan is an SME loan with an amount between 10,000 (20,000) and 500,000 USD
                       (1=yes, 0=no)
Requested amount       Requested loan amount (USD)
Collateral             Value of collateral as a share of the requested amount (%)
Repeat loan            Loan is a repeat loan vs. first loan (1=yes, 0=no)
Manat requested        Loan is requested in the local currency AZN (Manat) vs. USD (1=yes, 0=no)
Branch                 Branch dummies which are one if loan was granted at one of the following
                       branches: Baku, Gyanja, Khachmaz, Lenkoran, Sheki and Sumqayit
Petroleum location     Loan was granted in a location with high importance of the oil and gas sector, i.e.
                       Baku or Sumqayit (1=yes; 0=no)

Firm characteristics
Age                    Age of firm owner at date of disbursement of first loan (years)
Male                   Firm owner is male vs. female (1=yes, 0=no)
Married                Firm owner is married at date of disbursement (1=yes, 0=no)




                                                                                                             30
                                  Table 3. Descriptive statistics

                            Panel A. Loan and firm characteristics

This table displays summary statistics for our loan and firm variables based on the main analysis
sample starting in 2006. See Table 2 for definitions of all variables. Note that summary statistics for
the variable Approved are calculated including all loans while for all other variables only those loans
that were approved by the bank are included in the calculations.

                                   N            Mean          Median        Minimum        Maximum
Loan characteristics
Approved                          168,483            0.94              1              0              1
Real economy shock                118,848            0.55              1              0              1
Funding shock                      76,131            0.30              0              0              1
Agro                              158,481            0.18              0              0              1
Micro                             158,481            0.79              1              0              1
SME                               158,481            0.03              0              0              1
Repeat loan                       158,481            0.47              0              0              1
Requested amount                  158,481           3,986          2,000             80      5,000,000
Collateral                        158,481             275            192              0         34,618
Manat requested                   158,481            0.59              1              0              1

Firm characteristics
Age                               158,481              42              42            19             83
Male                              158,481            0.83               1             0              1
Married                           158,481            0.72               1             0              1




                                                                                                    31
                                                         Panel B. Sample means by loan type and crisis

This table reports sample means of our loan and firm variables for the non-crisis (years 2006 and 2007) and the Real economy shock period for the subsamples of
Agro, Micro and SME loans. ***, **, * denote that variables are significantly different from each other at the 0.01-, 0.05- and 0.1-level using a two-sided T-test.
See Table 2 for definitions of all variables. Note that summary statistics for the variable Approved are calculated including all loans while for all other variables
only those loans that were approved by the bank are included in the calculations.

                                             Agro loans                                    Micro loans                                   SME loans
                                Real           Real                             Real         Real                           Real           Real
                              economy        economy                          economy      economy                        economy        economy
                              shock = 0     shock = 1           Diff         shock = 0    shock = 1          Diff         shock = 0      shock = 1         Diff
                              N = 2,163     N = 18,366                       N = 49,022   N = 45,776                      N = 2,141      N = 1,380

Loan characteristics
Approved                            0.95           0.96      -0.02     ***         0.94           0.94    -0.01     **          0.89           0.84      0.06     ***
Repeat loan                         0.34           0.34          0                 0.40           0.56    -0.16     ***         0.70           0.73     -0.03     **
Requested amount                   1,710          2,180       -470     ***        2,271          3,109     -838     ***       41,333         51,307    -9,973     ***
Collateral                           343            240        251     ***          357            202      155     ***          501            462        40     **
Manat requested                     0.49           0.90      -0.41     ***         0.41           0.69    -0.28     ***         0.02           0.01      0.01

Firm characteristics
Age                                44.47          43.68       0.79     ***        42.25          40.97    1.28      ***        43.46          42.68      0.78     ***
Male                                0.88           0.90      -0.02     ***         0.83           0.80    0.03      ***         0.91           0.91         0
Married                             0.80           0.79       0.01                 0.72           0.70    0.03      ***         0.83           0.82      0.01
                            Table 4. Funding shock and loan approvals

This table reports results from LPM regressions for the full sample and the subsamples of Agro, Micro
and SME loans for the period 2006-2009. Standard errors are reported in parentheses and account for
clustering at the borrower level. All coefficients are multiplied by 100 so that they can directly be read
as percentages. The dependent variable is Approved which is a dummy variable indicating whether a
loan application was approved by the bank or rejected. All explanatory variables are defined in Table
2. ***, **, * denote significance at the 0.01-, 0.05- and 0.1-level

                                                   (1)           (2)              (3)           (4)
                                              Full sample     Agro loans      Micro loans     SME loans

Funding shock                                    2.946***            0.951       1.391***        7.027***
                                                    (1.125)        (0.826)         (0.347)         (2.485)
Agro                                             8.409***
                                                    (0.865)
Micro                                            5.000***
                                                    (0.726)
Agro*Funding shock                                 -2.251*
                                                    (1.252)
Micro*Funding shock                                -1.979*
                                                    (1.141)
Repeat loan                                      5.460***         2.299**        5.701***       10.831***
                                                    (0.170)         (0.938)        (0.206)          (1.471)
Repeat loan*Funding shock                                             0.546      -0.836**        -6.452**
                                                                    (1.091)        (0.396)          (2.713)
Ln(Requested amount)                            -0.579***        2.123***       -0.655***            -0.198
                                                   (0.136)          (0.495)        (0.144)          (0.763)
Collateral                                       0.006***        0.016***        0.006***        0.005***
                                                   (0.001)          (0.003)        (0.001)          (0.001)
Manat requested                                  1.250***         1.633**        1.222***          -9.182*
                                                   (0.193)          (0.712)        (0.201)          (5.170)
Age                                                 0.018*            0.008       0.024**            -0.081
                                                   (0.010)          (0.027)        (0.010)          (0.066)
Male                                             0.660***          -1.362*       0.895***        -3.359**
                                                   (0.238)          (0.725)        (0.253)          (1.637)
Married                                          1.245***         1.455**        1.153***         2.913**
                                                   (0.223)          (0.733)        (0.236)          (1.479)
Constant                                        86.619***       74.273***       91.778***       86.720***
                                                   (1.659)          (4.510)        (1.267)          (9.061)
Observations                                        80,277            6,303         70,765            3,209
R² adjusted                                          0.035            0.043          0.035            0.045
Branch fixed effects                                   yes              yes            yes              yes
                        Table 5. Real economy shock and loan approval

This table reports results from LPM regressions for the full sample and the subsamples of Agro, Micro
and SME loans for the period 2006-2009. Standard errors are reported in parentheses and account for
clustering at the borrower level. All coefficients are multiplied by 100 so that they can directly be read
as percentages. The dependent variable is Approved which is a dummy variable indicating whether a
loan application was approved by the bank or rejected. All explanatory variables are defined in Table
2. ***, **, * denote significance at the 0.01-, 0.05- and 0.1-level.

                                               (1)            (2)               (3)            (4)
                                          Full sample      Agro loans       Micro loans      SME loans

Real economy shock                           -5.403***         2.695***          0.678**        -6.998***
                                                (1.077)          (0.687)          (0.292)          (2.448)
Agro                                          6.760***
                                                (0.846)
Micro                                         3.999***
                                                (0.709)
Agro*Real economy shock                       6.555***
                                                (1.182)
Micro*Real economy shock                      5.518***
                                                (1.089)
Repeat loan                                   4.991***         2.717***         5.906***        10.325***
                                                (0.145)          (0.932)          (0.204)           (1.475)
Repeat loan*Real economy shock                                    -1.520       -1.311***              1.699
                                                                 (0.968)          (0.317)           (2.678)
Ln(Requested amount)                         -0.793***         1.427***        -0.866***             -0.807
                                                (0.124)          (0.355)          (0.133)           (0.808)
Collateral                                    0.006***         0.020***         0.006***         0.005***
                                                (0.001)          (0.002)          (0.001)           (0.001)
Manat requested                               1.359***         2.747***         1.246***       -14.507***
                                                (0.173)          (0.534)          (0.182)           (4.933)
Age                                              0.014*            0.007         0.023**           -0.109*
                                                (0.008)          (0.016)          (0.009)           (0.064)
Male                                              0.230       -1.543***          0.502**             -0.667
                                                (0.194)          (0.392)          (0.212)           (1.718)
Married                                       1.012***            0.743*        0.939***          3.166**
                                                (0.184)          (0.410)          (0.202)           (1.430)
Constant                                     89.869***        76.688***        93.750***        92.561***
                                                (1.518)          (3.085)          (1.155)           (9.581)
Observations                                   118,135            17,536           96,608             3,991
R² adjusted                                       0.031            0.040            0.028             0.054
Branch fixed effects                                yes              yes              yes               yes




                                                                                                       34
                                                            Table 6. Loan approvals and firm location

This table reports results from LPM regressions for the full sample and the subsamples of Agro, Micro and SME loans for the period 2006-2009. Standard errors
are reported in parentheses and account for clustering at the borrower level. All coefficients are multiplied by 100 so that they can directly be read as percentages.
The dependent variable is Approved which is a dummy variable indicating whether a loan application was approved by the bank or rejected. All explanatory
variables are defined in Table 2. ***, **, * denote significance at the 0.01-, 0.05- and 0.1-level.

                                   (1)          (2)          (3)          (4)                                           (5)          (6)          (7)          (8)
                                  Full         Agro         Micro        SME                                                        Agro         Micro
                                 sample        loans        loans        loans                                      Full sample     loans        loans      SME loans

Petroleum location               1.564***      2.609**     1.521***        6.614*   Petroleum location                1.587***      1.188**     1.480***          0.774
                                   (0.245)      (1.151)      (0.257)      (3.403)                                       (0.178)      (0.552)      (0.197)       (2.101)
Funding shock                    1.391***      1.572**     1.198***    12.078***    Real economy shock 2008            0.808**     1.824***         0.415         3.517
                                   (0.349)      (0.715)      (0.425)      (4.306)                                       (0.341)      (0.422)      (0.503)       (4.088)
                                                                                    Real economy shock 2009           0.973***     1.453***     0.859***    -11.766***
                                                                                                                        (0.220)      (0.291)      (0.324)       (3.579)
Funding shock*Petroleum                                                             Real economy shock
location                            -0.268       -1.366       -0.108   -10.596**    2008*Petroleum location          -1.584***    -3.423***      -1.092*         -3.732
                                   (0.394)      (1.365)      (0.466)      (4.464)                                       (0.437)      (1.251)      (0.576)       (4.356)
                                                                                    Real economy shock
                                                                                    2009*Petroleum location          -1.744***        -0.908   -1.324***          3.292
                                                                                                                        (0.269)      (0.681)      (0.361)       (3.793)
Agro                             6.558***                                                                             8.344***
                                   (0.709)                                                                              (0.558)
Micro                            4.472***                                                                             5.539***
                                   (0.650)                                                                              (0.542)
Constant                        86.223***    71.428***    91.285***    79.488***                                     86.744***    75.843***    93.324***    79.263***
                                   (1.628)      (4.421)      (1.260)      (9.086)                                       (1.303)      (2.508)      (0.993)      (7.974)
Observations                        80,277        6,303       70,765        3,209                                      158,840        24,984     128,497         5,359
R² adjusted                          0.028        0.039        0.027        0.043                                         0.024        0.035        0.021        0.051
Loan and firm characteristics          yes          yes          yes          yes                                           yes          yes          yes          yes
Branch fixed effects                    no           no           no           no                                            no           no           no           no
Figure 1. Timeline of events
         Figure 2. Loan applications and approvals for new vs. repeat borrowers

This figure displays the total number of loan applications and approvals for the three subsamples of
Agro, Micro and SME loans and for new vs. repeat borrowers respectively. The two vertical lines
indicate the start of the funding shock (April 2008) and the real economy shock (October 2008).

                                     Figure 2a. Agro loans




                                     Figure 2b. Micro loans




                                     Figure 2c: SME loans
                Figure 3. Loan approval rates for new vs. repeat borrowers

This figure displays approval rates for the three subsamples of Agro, Micro and SME loans and for
new vs. repeat borrowers respectively. The two vertical lines indicate the start of the funding shock
(April 2008) and the real economy shock (October 2008).

                                      Figure 3a. Agro loans




                                     Figure 3b. Micro loans




                                      Figure 3c. SME loans




                                                                                                  38
     Appendix 1. Exports 2006-2009


     Total exports (in thousand USD)




Exports to CIS countries (in thousand USD)




                                             39