220 Week 2 A /For WIZARD KIM
z06jl
out2.pdf
Do real estate loans reflect regional banking and
economic conditions? Amit Ghosh
Department of Economics, Illinois Wesleyan University, Bloomington, Illinois, USA
Abstract Purpose – Using state-level data, the purpose of this paper is to examine state banking-industry specific as well as region economic determinants of real estate lending of commercial banks across all 51 states spanning the period 1966-2014. Design/methodology/approach – Using both fixed-effects and dynamic-generalized method of moments (GMM) estimation techniques the study compares the sensitivity of different categories of real estate loans to regional banking and economic conditions. Finally, it provides a comparative perspective by comparing the results for real estate loans with other categories of loans given out by banks. Findings – Greater capitalization, liquidity and overhead costs reduce real estate lending, while banks diversification and the size of the banking industry in each state increase such lending. Moreover, real estate loans are found to be procyclical to state economic cycles with a rise in state real gross domestic product (GDP) growth, increase in state housing price index (HPI) and decline in both inflation and unemployment rates, increasing real estate loans. Within disaggregated loan types, construction and land development and single-family residential loans are most responsive to state banking and economic conditions. Originality/value – The recent financial turmoil is to a large extent attributable to excessive risk-taking by banks, particularly in terms of real estate lending. Hence, it is of paramount importance to empirically address the various determinants of real estate lending. With most banks restricting their operations in either one or a few states only, real estate lending in any given state may be more sensitive to regional banking and economic conditions than national aggregates. The present study is the first of its type to perform such an analysis.
Keywords Mortgages, Banks, Financial institutions and services, Models with panel data, Real estate services
Paper type Research paper
1. Introduction The US banking industry was at the center of the 2007-2009 financial crises that had deleterious consequences for banks’ financial health. Banks across the USA were hit by a sharp decline in their profitability along with an erosion of their capital cushions, which put severe pressure on their liquidity positions. These developments along with the overall poor health of the US economy imposed serious strains on banks’ balance sheet position and potentially impaired their ability to provide new loans. At the same
JEL classification – R10, R11, E32, G21, G28, C23 Comments by two anonymous referees and the Editors of the journal are gratefully
acknowledged.
The current issue and full text archive of this journal is available on Emerald Insight at: www.emeraldinsight.com/1757-6385.htm
Regional banking and
economic conditions
37
Received 11 September 2015 Revised 9 October 2015
Accepted 2 November 2015
Journal of Financial Economic Policy
Vol. 8 No. 1, 2016 pp. 37-63
© Emerald Group Publishing Limited 1757-6385
DOI 10.1108/JFEP-09-2015-0050
time, the origins of the recent financial turmoil are to a large extent attributable to excessive risk taking by banks, particularly in terms of real estate lending. In the build up to the crisis, concerns loomed amongst the federal banking regulatory agencies that concentration in commercial real estate loans has reached a level that could lead to undesirable outcomes in the event of a significant downturn. Such concerns became true from late 2008 onwards, with a precipitous decline in housing prices followed by large scale loans defaults, leading to a spat of bank failures, and the ensuing credit crunch that declined real estate lending (Lu and Whidbee, 2013; Rioja et al., 2014). This has sparked a burgeoning body of literature examining different aspects of research on bank lending, including real estate lending (Berrospide and Edge, 2010; Contessi and Francis, 2013; Ivashina and Scharfstein, 2010; Igan and Pinheirp, 2010; Peni et al., 2013). However, most studies use micro datasets and macro level empirical research is somewhat lacking. Pointedly, real estate loans are by far the largest loan category in the loan portfolios of most banks. Therefore, it is of paramount importance to empirically address the various determinants of real estate lending in the USA. Formal empirical research has also been very limited on the role of regional banking and economic conditions in affecting real estate loans. To the best of my knowledge, the present study is the first of its type to perform such an analysis.
Against this background, the focus of this paper is to examine the sensitivity of real estate loans to state-level macroeconomic conditions, while at the same time controlling for different state-level banking conditions. With this aim in mind, a panel econometric approach is used, encapsulating the largest time period of 1966-2014, and spanning across all 50 US states and District of Columbia. First, the real estate loans-elasticities with respect to both state-level economic as well as banking conditions are estimated. Thereafter, different categories of real estate loans data are used to calculate the impact of both state-level economic and banking variables, given different types of real estate loans are associated with different risk characteristics. Finally, a comparative perspective is provided by comparing the results for real estate loans with other categories of loans given out by banks.
The use of state-level data is motivated by the fact that the US commercial banking industry had restrictions on branching geographically due to its unique historical institutional origins. As a legacy of this, until today, most banks restrict their operations in either one or a few states only. Thus, bank lending in any given state may be more sensitive to regional conditions than national aggregates. Significant heterogeneity among banks across states also persists. Therefore, regional trends in real estate loan expansion and contraction may be increasingly sensitive to state macroeconomic conditions. The role of regional economic indicators in influencing real estate lending is further motivated by the fact that many states with large declines in house prices also experienced relatively large declines in personal income and gross state product and relatively large increases in unemployment rates (Depken et al., 2011). Hence, it remains interesting to examine the extent to which changes in real estate loans are causally associated with such changes in regional economic conditions across states. In general, the use of real estate as collateral lets businesses and consumers borrow more during regional economic booms (e.g. high state income growth and low inflation), which generally coincide with state real estate booms. As they borrow more, demand for real estate increases, pushing prices even higher and banks keep on lending. However, when the cycle starts turning (generally coinciding with decreasing or negative state income
JFEP 8,1
38
growth), banks find themselves with nonperforming loans and credit rationing sets in (Igan and Pinheirp, 2010).
From a policy perspective, understanding the relative significance of state-industry level versus regional economic factors in affecting real estate loans is extremely crucial for banks in making their investment decisions across the USA, which is characterized by stark state-wide heterogeneity. Furthermore, an important responsibility of the central bank or any banking supervisory authority is to ensure a stable and efficient banking system. It is in regard to prudential banking, supervision bank stress tests of loan quality are most useful (Barth et al., 2013; Marcelo et al., 2008). Therefore, from the perspective of restoring confidence in the US commercial banking industry, the findings of this study bear relevance for stress tests.
The rest of the paper is structured as follows. Section 2 provides trends and patterns in real estate lending across states in the USA. Section 3 theoretically discusses the different determinants of real estate lending and their theoretical underpinnings. Section 4 describes the data, and Section 5 presents the model and explains the results. Finally, section 6 provides policy implications of the findings.
2. Trends and patterns in commercial bank real estate lending in the USA Figure 1 shows the time series of real estate loans (in real terms) for 1966-2014. To provide a relative perspective, the series for the other categories of lending – agricultural loans, commercial and industrial (C&I) loans, individual loans and loans to other depository institutions are also plotted[1]. Real estate loans have grown substantially with a steep increase over the last three decades, especially from 1981 to 2014. From 1966 to 1986, C&I loans constituted the highest category, after which it was surpassed by real estate loans. The average annual growth rates over this entire time period was highest for real estate loans at 4.93 per cent followed by 3 per cent for individual loans, 2.61 per cent for C&I loans, 1.34 per cent for interbank loans and 0.55 per cent for agricultural loans. For years 2009, 2010 and 2011, real estate loans declined by 0.12, 5.95 and 5.77 per cent, respectively, from their previous years, reflecting the real estate bust in the USA. In terms of percentage shares, real estate loans had the highest share of all loans made by commercial banks across the USA from 1987 onwards. The share of real estate loans in banks overall loans portfolio stood at 55, 56 and 59 per cent for the years 2007-2009, respectively.
0
50,000,000
1E+09
1.5E+09
2E+09
2.5E+09
3E+09
3.5E+09
4E+09
19 66
19 70
19 74
19 78
19 82
19 86
19 90
19 94
19 98
20 02
20 06
20 10
20 14
Real Estate Loans
Loans to Depository Instittutions
Agricultural Loans
Commercial & Industrial Loans
Individual Loans
Figure 1. Time series of
different categories of loans
39
Regional banking and
economic conditions
It is important to examine real estate loans at a disaggregated level, as they differ in their characteristics, risk levels etc., which in turn affects bank lending patterns during different phases of business cycles. Figure 2 shows the ocular view of five different real estate loan categories – construction and land development, single-family residential, multi-family residential, farmland and non-farmland loans[2]. Single-family residential loans dominated real estate lending by commercial banks over the period of analysis followed by non-farm non-residential loans. For the period 1966-2014, on average 54.7 per cent of real estate loans were single-family residential followed by non-farm non-residential loans with an average share of 27.7 per cent. For the financial crises years 2007 to 2009, 54-56 per cent of real estate loans given by commercial banks across the nation were single-family residential loans. The share of multi-family residential loans was much smaller and stood between 4 and 5 per cent during these three years.
Next the distribution of real estate loans across states in the USA is dissected. Table I shows the average real estate loans given out by banks in each state over the entire time period. The states with the five highest average real estate loans were North Carolina, Ohio, California, New York and South Dakota. On the other end, the five states with the lowest average real estate loans were Idaho, Wyoming, Alaska, District of Columbia and New Hampshire. Real estate loans (in absolute terms) is expected to be higher in states with large size banking industry like California, New York, etc. Size is controlled for by scaling the amount of real estate loans by total assets in the banking industry of each state. This ratio was highest in Vermont followed by Oregon, Maryland, Wisconsin, New Hampshire and Connecticut; while, the five states where banks had the lowest share of real estate investments in their assets were New York, Delaware, Massachusetts, South Dakota and Nebraska.
Another key issue is how dominant are real estate loans in banks overall loans portfolio? Over exposure to the real estate sector along with deteriorating credit standards and risk management practices is often a concern among federal banking regulators. Column 3 answers this. Over the period of the study, banks in Vermont again had the highest average annual exposure to real estate loans followed by West Virginia, New Jersey, Connecticut, District of Columbia and Maryland. At the other extreme, the state with the lowest average share of real estate loans was South Dakota at 22 per cent, followed by Delaware, New York, Nebraska and Massachusetts.
Table II shows the averages of these variables for the top five states for the banking crisis years of 2007-2010. For a comparative perspective, the same are shown for
0
50,000,000
1E+09
1.5E+09
2E+09
2.5E+09
19 66
19 70
19 74
19 78
19 82
19 86
19 90
19 94
19 98
20 02
20 06
20 10
20 14
Construction and Land Development
1-4 Family Residential Properties
Multifamily Residential Properties
Farmland
Nonfarm Non- Residential
Figure 2. Time series of different categories of real estate loans
JFEP 8,1
40
Table I. Average real estate
loans, real estate loans-to-assets, real estate loans-to-total loans across states
States Real estate loansa Real estate
loans-to-total assets Real estate
loans-to-total loans
Alabama 37.600 0.282 0.452 Alaska 1.573 0.279 0.522 Arizona 6.117 0.272 0.390 Arkansas 9.366 0.293 0.507 California 102.000 0.307 0.462 Colorado 9.204 0.266 0.469 Connecticut 6.596 0.333 0.550 Delaware 28.500 0.138 0.240 District of Columbia 1.567 0.321 0.546 Florida 32.600 0.332 0.541 Georgia 40.200 0.267 0.412 Hawaii 5.707 0.310 0.521 Idaho 1.349 0.251 0.390 Illinois 53.100 0.201 0.363 Indiana 18.400 0.310 0.500 Iowa 10.200 0.242 0.417 Kansas 7.708 0.216 0.383 Kentucky 12.700 0.308 0.498 Louisiana 11.000 0.254 0.445 Maine 3.756 0.292 0.506 Maryland 11.300 0.346 0.539 Massachusetts 13.600 0.137 0.343 Michigan 27.400 0.312 0.488 Minnesota 19.000 0.277 0.431 Mississippi 9.746 0.282 0.482 Missouri 21.200 0.280 0.482 Montana 3.113 0.246 0.414 Nebraska 5.300 0.174 0.279 Nevada 28.300 0.204 0.336 New Hampshire 1.847 0.343 0.506 New Jersey 20.300 0.318 0.560 New Mexico 3.353 0.264 0.470 New York 78.900 0.117 0.251 North Carolina 177.000 0.223 0.391 North Dakota 2.726 0.211 0.348 Ohio 136.000 0.219 0.390 Oklahoma 9.609 0.218 0.396 Oregon 7.861 0.351 0.500 Pennsylvania 40.200 0.260 0.432 Pennsylvania 40.200 0.260 0.432 Rhode Island 18.400 0.293 0.449 South Carolina 8.394 0.307 0.485 South Dakota 88.600 0.142 0.224 Tennessee 20.300 0.295 0.472 Texas 42.800 0.208 0.373 Utah 12.600 0.198 0.332 Vermont 1.956 0.441 0.641
(continued)
41
Regional banking and
economic conditions
1987-1992, the period surrounding the Savings and Loans (S&L) crisis of the late 80s and early 90s. It reveals that the exposure of banks to real estate loans in their total loans portfolio as well as overall assets were higher for the top five states during the recent banking crisis years compared to the corresponding top five states over the earlier S&L crisis period.
3. Determinants of bank lending in the US banking industry The present study focuses on identifying the balance sheet constraints in determining loan growth, while at the same time controlling the impact coming from the demand side and other factors affecting banks’ lending behavior. It also carefully culls the theoretical debates on the impact of different banking and economic variables on real estate loans.
3.1 State banking-industry specific determinants of bank loans 3.1.1 Bank capitalization. The effect of bank capitalization on loans growth is ambiguous. On the one hand, well-capitalized banks or banks with access to additional sources of capital will be able to accommodate capital losses without reducing its assets, and, hence, its lending. By contrast, banks may very actively manage their assets to maintain a constant equity capital-to-assets ratio (Berrospide and Edge, 2010). Under this scenario, a capital loss results in a reduction in assets, and hence reduces real estate
Table I.
States Real estate loansa Real estate
loans-to-total assets Real estate
loans-to-total loans
Virginia 32.500 0.269 0.434 Washington 10.800 0.327 0.470 West Virginia 6.019 0.331 0.563 Wisconsin 22.800 0.339 0.531 Wyoming 1.410 0.246 0.446
Note: a In millions of US dollars
Table II. Top five states during two periods of banking crisis
Time-period
Average nominal real estate loans
Average real estate loans-to-total loans (%)
Average real estate loans-to-total assets (%)
2007-2010 North Carolina (714)a Oregon (87.1) Oregon (69.9) Ohio (525) District of Columbia (83.8) Arizona (60.9) South Dakota (280) New Hampshire (83.16) New Hampshire (59.86) Nevada (245) Vermont (82.16) Maryland (59.71) California (203) Florida (81.98) Washington (57.9)
1987-1992 California (107) Vermont (62.47) Vermont (47.6) New York (99.5) Florida (57.2) Maine (37.6) Florida (46.6) Connecticut (56.84) Connecticut (36.7) Pennsylvania (37.8) Maine (53.43) New Hampshire (35.9) Illinois (35.2) New Jersey (52.45) Florida (35.3)
Note: a In millions of US dollars
JFEP 8,1
42
lending because they cannot raise equity to offset declines in their capital. Moreover, higher capital buffers would suggest a more conservative management, which may undertake less risk, and hence reduce real estate lending (Skorepa and Seidler, 2015). Capitalization is measured by total equity capital-to-total assets.
3.1.2 Credit quality. Inferior credit quality is expected to reduce real estate lending. When banks credit quality deteriorates, banks focus on balance sheet repair and concentrate less on expanding loan availability. Hence, poor asset quality, a proxy for risk, should act as a drag on real estate loan growth. An alternative and potentially more dangerous scenario arises when banks with poor credit quality, instead of undertaking balance sheet repair, focus on lending expansion in a gamble for redemption (Dages et al., 2000). In that situation, poor credit quality will increase real estate lending. Credit quality is measured by provision for loan and lease losses-to-total loans.
3.1.3 Operating efficiency. The effect of cost efficiency is ambiguous. On the one hand, following the “skimping hypothesis” of Berger and DeYoung (1997) banks which devote fewer resources to monitor lending risks will be more cost-efficient. This implies a negative effect of efficiency on real estate lending. On the contrary, higher cost inefficiency would increase lending following the “bad management” hypothesis of Berger and DeYoung (1997), as bank managers with poor skills in credit scoring and monitoring borrowers increase costs and give out poor quality loans. Operating efficiency is measured by non-interest expenses divided by total assets (OCA).
3.1.4 Bank diversification. Banks income or earning streams can be decomposed into interest and non-interest incomes. The former includes traditional commercial bank activities like interest earned from different types of loans, and investment securities. The latter covers investment banking, asset management and insurance underwriting, fee-paying and commission-paying services, trading and derivatives. Diversification is measured by the share of non-interest income to total income for banks in each state. More diversification in the banks business model provides an additional avenue of earnings for banks that could be used to increase real estate lending. Thus, a positive impact of bank diversification on lending is expected, capturing the complementary relationship between non-interest income and interest-based earnings such as real estate loans.
From another perspective, the share of non-interest income in total income is used as a more forward-looking measure of risk. The higher is the share of non-interest income, the higher the volatility of returns, and thus higher the risk. Existing empirical work suggests that non-interest income generating activities are substantially riskier than traditional credit business (Buch et al., 2014 and the references cited therein). Assuming that banks were aware of the risks associated with these investments, an increase in the non-interest income ratio can be interpreted as evidence of risk taking by banks and should reduce banks’ lending.
3.1.5 Liquidity risks. Banks with a higher share of more liquid assets are perceived to have a safer asset portfolio, hence would engage in undertaking less riskier investments. Thus, greater liquidity (and hence lower liquidity risks) should reduce bank lending. Liquidity is measured by the sum of cash and investment securities-to-total assets. This variable is intended to capture the extent to which banks use their stock of securities to adjust their loan growth (Kashyap and Stein, 2000).
3.1.6 Industry size. Banks in states with large-sized banking industry can increase their leverage too much and extend greater loans, particularly to lower quality
43
Regional banking and
economic conditions
borrowers. In large-sized markets, banks often resort to excessive risk taking because it is difficult to impose market discipline by regulators and banks expect government protection in the case of failures (Stern and Feldman, 2004). Thus, one expects more real estate lending in states with a larger banking industry. Moreover, banks in different sized markets have very different characteristics, business strategies and product compositions. Bank size may also surrogate for such variables. Following Laeven and Levine (2009), size is measured by the logarithm of total assets divided by the number of banks in each state.
3.1.7 Bank profitability. Highly profitable banks have fewer incentives to engage in high-risk activities. Thus, profitability is expected to negatively impact real estate lending, following the “bad management” hypothesis of Berger and DeYoung (1997). In rebuttal, higher profits could also increase lending. This possibility is shown in the model of Rajan (1994) where credit policy is not determined solely by the maximization of banks’ earnings but also by the short-term reputation concerns of banks’ management. Consequently, bank managers may attempt to manipulate current earnings resorting to a “liberal credit policy”. In this manner, a bank may attempt to convince the market for its profitability by inflating current earnings at the expense of increasing their loans. I measure profits by return on assets (ROA) of banks in each state.
3.2 Regional economic conditions 3.2.1 State GDP. Loan supply and demand may differ across states for numerous reasons. Bank lending may be distinct from one state to another in terms of lending motives with respect to their clients. Through “transaction-based” lending motives, improved economic conditions generate opportunities for expanding production and investment. Bank loans expand to accommodate part of this demand. Thus, transaction-based lending is procyclical (Dages et al., 2000). Moreover, according to the ability to pay hypothesis, borrowers will default on the loan when they face an income shock or an unfavorable change in loan terms that makes it implausible to keep up with their payments. While shocks related to the borrower’s personal circumstances (divorce, emergency medical care, etc.) tend to lead to defaults, it is the macro shocks such as rising interest rates or sluggish economic activity that are more likely to have a stronger relation with defaults at the aggregate level (Igan and Pinheirp, 2010).
In rebuttal, through “relationship-based” lending motives, bank lending helps established customers smooth the effects of cyclical fluctuations on consumption. Under adverse economic conditions, lending expands to offset some of the revenue shortfall of clients; under economic expansions, net lending by banks declines as borrowers pay back outstanding loans. Under these stylized conditions, bank lending is countercyclical to regional economic conditions (Dages et al., 2000). State-level GDP growth is used to control for changes in loan demand.
3.2.2 Regional inflation rates. The relationship between banks real estate lending and inflation is ambiguous. Theoretically, for unchanged nominal interest rates, inflation should reduce the real value of debt and, hence, make debt servicing easier. This should increase lending. However, high inflation may pass through to nominal interest rates, reducing borrowers’ loan-servicing capacity or it can negatively affect borrowers’ real income when nominal wages are sticky. If the income does not increase in line with inflation, a rise in inflation increases costs (for both households and corporates) and lowers the amount of available funds for debt repayment (Louizis et al., 2012; Nkusu,
JFEP 8,1
44
2011). This would reduce bank lending. State-level inflation data are not available. As its closest proxy, the percentage change of the consumer price index (CPI) of the largest urban center either in the state or closest to that state is used.
3.2.3 Real interest rates. A rise in real lending rates (i.e. with floating interest rates) increases the real value of borrowers’ debt and makes debt servicing more expensive. This will increase loan defaults and reduce bank lending. The 30-year mortgage rate is used to proxy interest rates.
3.2.4 State housing price index. Rising home prices boost financial wealth and can help borrowers face unexpected adverse shocks or ease their access to credit by boosting the value of the underlying homes used as collateral (Beck et al., 2015; Ghosh, 2015; Nkusu, 2011). In this regard, changes in housing price are expected to positively affect bank lending. From another perspective, strategic-default hypothesis, argues that borrowers will default when the value of the loan is greater than the value of the asset, after taking transaction and reputation costs into account. Especially with collaterized loans, incentives to exercise the option to default will depend on the movements in the underlying asset value. Depressed house prices could, especially, trigger defaults on residential mortgages and reduce real estate loans (Igan and Pinheirp, 2010).
3.2.5 State unemployment rates. Rising unemployment typically causes more loan defaults, thereby lowering bank profits and hence reducing bank lending.
4. Data description and preliminary statistical diagnostics Banking-industry specific data are retrieved from the balance sheet and income statements of each state. These are available on the federal deposit insurance corporation (FDIC) website under “Historical Statistics on Banking”. State GDP and personal income are sourced from US bureau of economic analysis (BEA), while state-level unemployment rates and regional CPI are taken from US Bureau of Labor Statistics (BLS). Finally, state housing price indices are taken from the US Federal Housing Finance Agency and interest rate data are sourced from the Federal Reserve Board. The data span from 1966 to 2014, except for state unemployment rates that are available from 1976 and state HPI that commence in 1975[3]. Table III summarizes the variables and their corresponding sources.
Using state-level data is motivated by the fact that most micro studies rely on cross-section dimension of the data more than the time dimension. This raises doubts on the accuracy of the estimates when the impact of macroeconomic variables is in question because of little, if any, variation in the macro variables used. Exploiting cross-state variation in bank lending trends is also likely to yield more robust results than an analysis of bank-level data or individual states, because time series for loans are typically short. Moreover, aggregate data for the entire banking system of each state (in contrast to bank-level data) are considered preferable as the risk of non-representativeness of the sample is reduced (Boudriga et al., 2009; Igan and Pinheirp, 2010).
4.1 Panel unit root tests Real estate and other categories of loans are expressed in growth rates to deal with non-stationary variables. All other variables are also expressed in their logarithmic forms. The Levin et al. (2002) panel unit root test that assumes a common unit root process is performed, and the Im et al. (2003) Fisher-ADF test that assume individual
45
Regional banking and
economic conditions
Table III. Description of variables
V ar
ia bl
es D
es cr
ip ti
on D
at as
ou rc
e M
ea n
SD N
L og
(r ea
lr ea
le st
at e
lo an
s) R
ea le
st at
e lo
an s
as ev
id en
ce d
by m
or tg
ag es
,d ee
ds ,l
an d
co nt
ra ct
s, et
c. F
D IC
16 .2
70 1.
37 6
2, 49
9
L og
(r ea
lC &
I lo
an s)
L oa
ns to
so le
pr op
ri et
or s,
pa rt
ne rs
hi ps
,c or
po ra
ti on
s an
d ot
he r
bu si
ne ss
es F
D IC
15 .0
13 1.
60 4
2, 49
9
L og
(r ea
la gr
ic ul
tu ra
ll oa
ns )
L oa
ns m
ad e
to fi
na nc
e ag
ri cu
lt ur
al pr
od uc
ti on
an d
ot he
r lo
an s
to fa
rm er
s F
D IC
12 .8
00 2.
21 1
2, 47
0
L og
(r ea
ll oa
ns to
de po
si to
ry In
st it
ut io
ns )
L oa
ns to
ba nk
s an
d ot
he r
de po
si to
ry in
st it
ut io
ns F
D IC
11 .4
73 2.
62 2
2, 36
0 L
og (r
ea li
nd iv
id ua
ll oa
ns )
L oa
ns fo
r ho
us eh
ol d,
fa m
ily an
d ot
he r
pe rs
on al
ex pe
nd it
ur es
F D
IC 15
.3 39
1. 58
0 2,
49 9
L og
(r ea
lc on
st ru
ct io
n an
d la
nd de
ve lo
pm en
t lo
an s)
L oa
ns to
fi na
nc e
la nd
de ve
lo pm
en t
or co
ns tr
uc ti
on of
fa rm
bu ild
in gs
F D
IC 14
.1 93
1. 51
8 1,
98 9
L og
(r ea
lf ar
m la
nd lo
an s)
L oa
ns se
cu re
d by
fa rm
la nd
,i nc
lu di
ng la
nd fo
r ag
ri cu
lt ur
al pu
rp os
es F
D IC
12 .3
01 2.
01 4
2, 48
7
L og
(r ea
l1 -4
fa m
ily re
si de
nt ia
lp ro
pe rt
ie s
lo an
s) L
oa ns
on 1-
4 fa
m ily
dw el
lin g
un it
s, m
ob ile
ho m
es ,i
nd iv
id ua
l co
nd os
et c
F D
IC 15
.5 51
1. 42
4 2,
49 9
L og
(r ea
lm ul
ti fa
m ily
re si
de nt
ia l
pr op
er ti
es lo
an s)
L oa
ns on
fi ve
or m
or e
dw el
lin g
un it
s in
ap ar
tm en
ts ,h
ou se
ke ep
in g
dw el
lin gs
,e tc
F D
IC 12
.7 76
1. 53
7 2,
34 6
L og
(r ea
ln on
-f ar
m la
nd lo
an s)
L oa
ns on
bu si
ne ss
an d
in du
st ri
al pr
op er
ti es
,h ot
el s,
m ot
el s,
ho sp
it al
s, cl
ub s
et c
F D
IC 15
.0 73
1. 33
7 2,
49 9
L og
(c ap
it al
-t o-
as se
t) T
ot al
eq ui
ty ca
pi ta
l-t o-
as se
ts F
D IC
� 2.
50 2
0. 24
0 2,
49 8
L og
(c re
di t
qu al
it y)
L oa
ns an
d le
as e
lo ss
pr ov
is io
n- to
-t ot
al lo
an s
F D
IC �
5. 31
6 0.
93 6
2, 43
2 L
og (d
iv er
si fi
ca ti
on )
N on
-in te
re st
in co
m e-
to -t
ot al
in co
m e
F D
IC �
1. 47
8 0.
38 8
2, 49
9 L
og (O
C A
) N
on -in
te re
st ex
pe ns
es -t
o- to
ta lA
ss et
s F
D IC
� 3.
49 1
0. 25
2 2,
49 9
L og
(R O
A )
N et
in co
m e-
to -t
ot al
as se
ts F
D IC
� 4.
68 7
0. 47
0 2,
39 5
L og
of in
fl at
io n
P er
ce nt
ch an
ge in
re gi
on al
C P
I U
S B
L S
0. 04
1 0.
02 8
2, 49
3 L
og (r
ea ll
oa n
ra te
) N
om in
al ra
te ch
ar ge
d on
lo an
s to
bu si
ne ss
es ad
ju st
ed fo
r re
gi on
al in
fl at
io n
F ed
er al
re se
rv e
0. 03
5 0.
02 4
2, 49
3
L og
(r ea
lF F
R )
N om
in al
ra te
ch ar
ge d
on in
te rb
an k
lo an
s ad
ju st
ed fo
r re
gi on
al in
fl at
io n
F ed
er al
re se
rv e
0. 01
4 0.
02 5
2, 49
3
L og
(r ea
lm or
tg ag
e ra
te )
In fl
at io
n- ad
ju st
ed in
te re
st ra
te s
fo r
30 -y
ea r
fi xe
d- ra
te fi
rs t
m or
tg ag
es F
ed er
al re
se rv
e 0.
04 1
0. 02
7 2,
19 2
R ea
ls ta
te G
D P
gr ow
th In
fl at
io n
ad ju
st ed
pe r
ce nt
ch an
ge in
G D
P U
S B
E A
2. 63
8 4.
07 4
2, 49
9 R
ea ls
ta te
pe rs
on al
in co
m e
gr ow
th In
fl at
io n
ad ju
st ed
pe r
ce nt
ch an
ge in
pe rs
on al
in co
m e
U S
B E
A 3.
21 6
22 .4
06 2,
49 9
L og
(h ou
si ng
pr ic
e in
de x)
St at
e- le
ve lH
P I
F H
F A
5. 19
9 0.
54 7
2, 04
0 L
og (U
ne m
pl oy
m en
t ra
te s)
U ne
m pl
oy ed
-t o-
la bo
r fo
rc e
in ea
ch st
at e
U S
B L
S 1.
75 3
0. 33
7 1,
98 9JFEP
8,1
46
Table IV. Panel unit root
results
V ar
ia bl
es L
ev in
,L in
an d
C hu
t P
ro ba
bi lit
y Im
,P es
ar an
an d
Sh in
W -s
ta t
P ro
ba bi
lit y
A D
F –
F is
he r
ch i-s
qu ar
e P
ro ba
bi lit
y
L og
(r ea
lr ea
le st
at e
lo an
s) �
0. 95
4 0.
17 0
� 0.
12 6
0. 45
0 10
0. 40
4 0.
52 6
L og
(r ea
lC &
I lo
an s)
� 3.
07 3
0. 00
1 2.
56 1
0. 99
5 60
.8 32
1. 00
0 L
og (r
ea la
gr ic
ul tu
ra ll
oa ns
) 0.
09 4
0. 53
7 0.
07 5
0. 53
0 95
.6 98
0. 65
7 L
og (r
ea ll
oa ns
to de
po si
to ry
in st
it ut
io ns
) �
2. 06
1 0.
02 0
� 1.
72 5
0. 04
2 13
0. 51
0 0.
03 0
L og
(r ea
li nd
iv id
ua ll
oa ns
) �
4. 03
8 0.
00 0
0. 76
9 0.
77 9
85 .0
05 0.
88 8
L og
(r ea
lc on
st ru
ct io
n an
d la
nd de
ve lo
pm en
t lo
an s)
� 3.
07 2
0. 00
1 �
4. 58
2 0.
00 0
16 2.
59 1
0. 00
0 L
og (r
ea lf
ar m
la nd
lo an
s) �
1. 82
0 0.
03 4
� 1.
38 8
0. 08
3 12
0. 00
2 0.
10 8
L og
(r ea
l1 -4
fa m
ily re
si de
nt ia
lp ro
pe rt
ie s
lo an
s) �
1. 05
5 0.
14 6
1. 23
9 0.
89 2
89 .2
28 0.
81 3
L og
(r ea
lm ul
ti fa
m ily
re si
de nt
ia lp
ro pe
rt ie
s lo
an s)
� 1.
88 6
0. 03
0 �
4. 44
9 0.
00 0
17 4.
13 9
0. 00
0 L
og (r
ea ln
on -f
ar m
la nd
lo an
s) 1.
16 3
0. 87
8 �
0. 18
8 0.
42 6
96 .7
43 0.
62 8
L og
(c ap
it al
-t o-
as se
t) �
6. 49
9 0.
00 0
� 3.
66 8
0. 00
0 14
6. 34
0 0.
00 3
L og
(c re
di t
qu al
it y)
� 0.
35 9
0. 36
0 �
5. 71
2 0.
00 0
19 8.
81 3
0. 00
0 L
og (d
iv er
si fi
ca ti
on )
3. 28
1 1.
00 0
1. 89
5 0.
97 1
79 .6
58 0.
95 0
L og
(O C
A )
� 1.
22 5
0. 11
0 1.
64 8
0. 95
0 86
.2 88
0. 86
8 L
og (R
O A
) 0.
32 0
0. 62
5 �
5. 37
1 0.
00 0
30 0.
98 5
0. 00
0 L
og of
in fl
at io
n �
23 .6
60 0.
00 0
� 16
.9 96
0. 00
0 46
3. 29
4 0.
00 0
L og
(r ea
ll oa
n ra
te )
2. 79
2 0.
99 7
� 0.
91 4
0. 18
0 88
.4 15
0. 82
9 L
og (r
ea lF
F R
) 0.
53 0
0. 70
2 �
3. 48
5 0.
00 0
12 8.
95 3
0. 03
7 L
og (r
ea lm
or tg
ag e
ra te
) �
8. 90
2 0.
00 0
� 4.
56 8
0. 00
0 14
9. 67
3 0.
00 2
R ea
ls ta
te G
D P
gr ow
th �
21 .6
65 0.
00 0
� 20
.1 95
0. 00
0 56
4. 63
3 0.
00 0
R ea
ls ta
te pe
rs on
al in
co m
e gr
ow th
� 19
.6 23
0. 00
0 �
17 .7
22 0.
00 0
53 4.
52 5
0. 00
0 L
og (H
P I)
� 6.
20 4
0. 00
0 �
6. 25
9 0.
00 0
21 6.
26 0
0. 00
0 L
og (u
ne m
pl oy
m en
t ra
te s)
� 8.
16 1
0. 00
0 �
9. 31
2 0.
00 0
26 3.
43 8
0. 00
0
47
Regional banking and
economic conditions
unit root processes on these other variables. Table IV presents the panel unit roots results, where the null hypothesis is of non-stationarity. Variables that exhibit unit roots in their levels form were first-differenced to induce stationarity[4].
4.2 Correlations Table V shows the pair wise correlation matrix among the variables. Real estate loans are highly positively correlated with size and, to a lesser magnitude, negatively correlated with overhead costs-to-assets and liquid assets-to-total assets. State banking industry size is negatively correlated with both OCA and liquidity, suggesting some concerns of multicollinearity. Turning to the correlations with the state economic variables shown in the lower panel, real estate loans are positively associated with both state GDP growth and changes in state HPI. The latter is positively correlated with state real GDP growth suggesting that both regional economic and real estate cycles tend to coincide together.
5. Econometric model, results and discussion 5.1 Real estate loan estimation Loan growth is modeled as a function of supply and demand factors, following Berrospide and Edge (2010), Peni et al. (2013). Specifically, the fixed-effects regression that is estimated takes the following form:
Yit � a0it � ajit(Xit j ) � akit (Xit
k) � �i � �it (1)
Where Yit denotes the growth rate of loans for state i in period t; (Xit j ) denotes a vector of
state banking-industry specific variables; and (Xit k) represents the vector of state
economic variables. i represent each state and t each year; � refers to state-fixed effects
Table V. Correlation matrix
Banking-industry variables
Real estate loans
Capital- to-assets
Credit quality Diversification OCA Liquidity Size ROA
Real estate loans 1 Capital-to-assets �0.091 1 Credit quality �0.050 0.054 1 Diversification 0.029 �0.065 0.095 1 OCA �0.224 �0.071 0.124 0.381 1 Liquidity �0.334 0.072 0.182 �0.047 0.066 1 Size 0.812 �0.100 �0.021 0.064 �0.237 �0.216 1 ROA 0.062 0.294 �0.176 0.036 �0.105 �0.185 0.002 1
state-economic variables
Real estate loans
Real GDP growth HPI Inflation
Unemployment rates
Real estate loans 1 Real GDP growth 0.163 1 HPI 0.174 0.318 1 Inflation �0.012 �0.075 0.250 1 Unemployment rates �0.131 �0.233 �0.244 0.018 1
JFEP 8,1
48
and �it is an independently and identically distributed error term. The static framework uses a fixed-effects estimation model that controls for the effect of time-invariant unobserved heterogeneity across states, captured by state-specific dummies. Because the regression analysis is limited to a specific set of states and all variables are time varying, it is reasonable to use this estimation technique as one of the methods. Moreover, the use of state-specific effects addresses the omitted-variables bias problem. The results are shown in the left panel of Table VI[5].
Both increases in overhead costs-to-assets and liquidity consistently reduce real estate lending. The results for the banks cost efficiency measure provide evidence of the “skimping hypothesis” with a 1 per cent decline in costs increasing real estate lending from 0.06 to 0.7 per cent. Likewise, as banks increase their share of liquid assets and lower liquidity risk, they refrain from real estate lending with a 1 per cent rise in liquidity reducing real estate lending by 0.27-0.55 per cent. The size of the banking industry is positively significant, implying more real estate loans are given out in states with a large-sized banking industry, with a 1 per cent increase in industry size, increasing lending by 0.92 per cent. However, as noted previously, this variable affects other banking-industry variables due to multicollinearity issues. When the size variable is dropped, banks capital-to-assets, credit quality and extent of diversification are found to be statistically significant. A 1 per cent rise in banks’ capital reduces real estate lending by 0.11-0.13 per cent. The negative coefficient implies conservative management practices by bank executives. Banks with higher capital-to-asset ratios are considered relatively safer and less risky compared to institutions with lower capital ratios. This result is consistent with the effect of capital regulatory requirements on bank fragility in Lee and Lu (2015). In addition, more capitalized banks have a high franchise value. Therefore, following conventional risk-return hypothesis engage in less risky lending. A 1 per cent rise in credit quality, signifying inferior quality of credit, and hence more credit risk, increases real estate lending by approximately 0.02 per cent. An increase in banks diversification by 1 per cent increases real estate lending by 0.26-0.29 per cent. This implies that an increase in the banks earning from non-interest-based activities positively contributes to real estate lending. Finally, banks ROA are largely insignificant except in Model 4, which suggests that as banks profitability rises, they reduce real estate lending following the “bad management” hypothesis.
Turning to the state economic variables, the results consistently illustrate an increase in state real GDP growth to significantly increase real estate lending, lending support to the “transaction-based” lending motives. A rise in unemployment rates lower real estate loans. These results imply real estate loans are procyclical to state economic cycles[6]. Inflation rates reduce real estate lending. This suggests that rise in nominal incomes may be lower than inflation rates making debt servicing difficult, and hence acts a drag on real estate loans. A decline in mortgage rates by 1 per cent makes borrowing cheaper and, hence, increases real estate loans by 0.47 per cent. Lastly, a rise in state-level housing price index by 1 per cent positively and significantly increases real estate loans by 0.55 per cent, again capturing its procyclical nature.
The state banking-industry specific variables could be endogenous with loan growth and with one another. To deal with such endogeneity concerns, the systems-GMM estimation developed by Arellano and Bover (1995) and Blundell and Bond (1998) is used. This methodology essentially regresses levels and changes in real estate loans on the lags of the same variable as well as other explanatory variables using lagged levels
49
Regional banking and
economic conditions
Table VI. Results for real estate loans
V ar
ia bl
es F
ix ed
ef fe
ct s
es ti
m at
io n
M od
el 1
M od
el 2
M od
el 3
M od
el 4
C on
st an
t 0.
00 5
(0 .1
06 )
� 0.
00 4
(� 0.
07 2)
� 0
.1 9
4 **
(� 2
.2 9
3 )
� 0 .2
3 9 **
(� 2 .3
8 8 )
C ap
it al
-t o-
as se
ts �
0. 00
9 (�
0. 71
) �
0. 01
2 (�
0. 84
) �
0 .1
0 5
** *
(� 4
.4 7
9 )
� 0 .1
3 **
* (�
4 .3
6 2 )
C re
di t
qu al
it y
0. 00
5 (1
.5 63
) 0.
00 4
(0 .9
98 )
0 .0
1 4
** (2
.0 3
6 )
0 .0
1 8 **
(2 .3
7 4 )
D iv
er si
fi ca
ti on
� 0.
01 2
(� 0.
53 8)
� 0.
01 6
(� 0.
69 2)
0 .2
5 5
** *
(6 .6
4 7
) 0 .2
8 6 **
* (7
.1 7 8 )
O C
A �
0 .0
7 **
(� 2
.2 7
4 )
� 0.
05 1
(� 1.
48 7)
� 0
.6 5
** *
(� 1
2 .1
2 6
) �
0 .6
6 8 **
* (�
1 1 .9
6 8 )
L iq
ui di
ty �
0 .2
8 **
* (�
1 3
.0 8
5 )
� 0
.2 7
1 **
* (�
1 1
.8 3
9 )
� 0
.5 4
9 **
* (�
1 4
.8 5
7 )
� 0 .5
3 **
* (�
1 3 .8
3 8 )
Si ze
0 .9
1 3
** *
(6 1
.2 5
6 )
0. 92
2* **
(5 8.
15 4)
R O
A 0.
01 0
(1 .4
45 )
0. 01
1 (1
.4 81
) �
0. 00
5 (�
0. 41
8) �
0 .0
3 3
(� 2 .5
1 1 )
R ea
lG D
P gr
ow th
0 .4
7 4
** *
(6 .4
7 )
0 .4
4 1
** *
(5 .4
2 7
) 0
.6 1
7 **
* (4
.6 0
6 )
0 .3
0 9 **
* (1
.8 7 3 )
R ea
lm or
tg ag
e ra
te s
� 0.
46 1*
** (�
2. 86
7) �
0. 31
1 (�
1. 17
1) In
fl at
io n
� 0 .8
** *
(� 3 .3
2 6 )
H P
I 0 .5
4 9 **
* (5
.7 6 7 )
U ne
m pl
oy m
en t
ra te
s �
0 .0
6 6 **
* (�
2 .9
6 1 )
y ( t�
1 )
A dj
us te
d R
2 or
C hi
-s qu
ar e
0. 69
6 0.
70 2
0. 18
8 0.
20 7
F -s
ta ti
st ic
92 .8
28 **
* 81
.6 31
** *
9. 08
4* **
9. 15
** *
D -W
1. 79
2 1.
81 9
1. 90
7 1.
94 8
N 2,
32 7
2, 02
2 2,
02 2
1, 86
9 A
R (1
) p-
va lu
es A
R (2
) p-
va lu
es (c
on ti
nu ed
)JFEP 8,1
50
Table VI.
V ar
ia bl
es Sy
st em
s- G
M M
es ti
m at
io n
M od
el 1
M od
el 2
M od
el 3
M od
el 4
C on
st an
t 0.
07 7
(1 .4
0) 0.
08 7
(1 .2
3) �
0. 00
9 (�
0. 07
) �
0. 07
5 (�
0. 53
) C
ap it
al -t
o- as
se ts
0 .0
2 5
* (1
.6 8
) 0.
02 3
(1 .5
9) �
0. 04
8 (�
1. 22
) �
0. 08
2* (�
1. 67
) C
re di
t qu
al it
y 0.
01 0
(1 .3
1) 0.
00 9
(1 .0
3) 0.
01 9
(1 .5
3) 0.
02 0
(1 .4
7) D
iv er
si fi
ca ti
on 0.
01 9
(0 .2
0) 0.
00 8
(0 .0
8) 0
.2 2
2 **
* (4
.0 8
) 0.
26 **
* (4
.9 1)
O C
A �
0. 09
8 (�
1. 15
) �
0. 06
0 (�
0. 64
) �
0 .6
7 5
** *
(� 4
.6 2
) �
0. 69
8* **
(� 4.
53 )
L iq
ui di
ty �
0 .3
0 2
** *
(� 2
.6 5
) �
0 .2
9 1
** *
(� 2
.4 9
) �
0 .5
8 2
** *
(� 2
.6 1
) �
0 .5
6 2 **
* (�
2 .4
4 )
Si ze
0 .9
3 4
** *
(1 4
.8 2
) 0.
94 4*
** (1
4. 95
) R
O A
0. 00
2 (0
.2 4)
� 0.
00 4
(� 0.
32 )
� 0.
02 6
(� 2.
21 )
R ea
lG D
P gr
ow th
0 .4
4 **
* (3
.3 0
) 0
.3 9
3 **
* (2
.7 7
) 0
.4 3
9 **
* (2
.7 7
) 0.
13 3
(0 .8
0) R
ea lm
or tg
ag e
ra te
s �
0. 50
4* **
(� 4.
28 )
� 0.
31 6*
(� 1.
83 )
In fl
at io
n �
0. 86
9* **
(� 4.
52 )
H P
I 0.
49 7*
** (3
.3 0)
U ne
m pl
oy m
en t
ra te
s �
0. 05
6 (�
1. 78
) y (
t� 1 )
0. 01
6 (0
.8 1)
0. 00
7 (0
.3 6)
� 0.
01 0
(� 0.
27 )
� 0.
02 9
(� 0.
78 )
A dj
us te
d R
2 or
C hi
-s qu
ar e
11 65
.4 6
11 64
.8 1
27 5.
56 28
9. 97
F -s
ta ti
st ic
D -W
N 2,
21 5
1, 96
1 1,
96 1
1, 80
8 A
R (1
) �
3. 73
7* **
� 3.
69 9*
** �
3. 27
3* **
� 3.
32 4*
** p-
va lu
es 0.
00 0
0. 00
0 0.
00 1
0. 00
1 A
R (2
) �
1. 43
4 �
1. 37
4 �
1. 59
4 �
1. 59
5 p-
va lu
es 0.
13 1
0. 12
0. 11
1 0.
11 1
N o te
s: T
er m
s in
br ac
ke ts
de no
te t-
st at
is ti
c ba
se d
on ro
bu st
st an
da rd
er ro
rs cl
us te
re d
in st
at es
; *
, **
, **
* in
di ca
te s
si gn
ifi ca
nc e
at th
e 10
,5 an
d 1%
le ve
l; bo
ld va
lu es
de no
te th
e st
at is
ti ca
lly si
gn ifi
ca nt
co ef
fi ci
en ts
51
Regional banking and
economic conditions
as instruments. This reduces potential biases in finite samples and any asymptotic imprecision:
Yit � a � �Yi, t�1 � � j�1 J
(ajXit j ) � � k�1
K (akXit
k) � �it (2)
The right panel of Table VI presents the results. The GMM estimations largely confirm the fixed-effects result with both state banking as well as economic variables showing the same signs and significance, with the exception of credit quality that is now statistically insignificant. The lagged loan coefficient is insignificant, suggesting no persistence in real estate loans.
5.2 Results for disaggregated categories of real estate loans Different types of real estate loans are generally associated with different risk characteristics, with single-family residential loans typically considered as the least risky and construction and land development loans as the most risky. In general, commercial real estate loans are further considered to be riskier than residential real estate loans, not only because the primary source of repayment is cash flows from the real estate collateral but also because commercial real estate prices have historically shown more volatility (Igan and Pinheirp, 2010)[7]. Due to these issues, the sensitivity of different real estate loans types to regional banking and economic conditions could differ. Tables VII and VIII present the results for the fixed-effects and GMM estimations, respectively[8].
The results using both estimation techniques are very similar. Comparing across the five categories of real estate loans, construction and land development are most sensitive to state banking and regional economic conditions, followed by single-family residential loans. Greater capitalization reduces the most risky construction and land development lending. The same applies for single-family residential loans. However, increase in capital causes modest increases in multifamily residential (only in the GMM estimations) and farmland loans and has no significant impact on non-farmland loans. As banks credit quality deteriorates, it discourages the most-risky construction and land development lending. However, it does increase other types of real estate lending. An increase in banks overhead costs reduce lending across all categories with the elasticities highest for single-family residential loans. Banks diversification significantly increases lending across real estate loan types, except for farmland loans, and the elasticities are highest again for single-family residential loans. Likewise, single-family residential loans are most sensitive to liquidity risks, with a 1 per cent rise in liquidity lowering such loans by 0.76-82 per cent. Non-farmland, non-residential loans are not significantly affected by banks liquidity. Bank profitability significantly reduces both single family- and multifamily residential loans. However as banks profits rise, it enhances risky construction and land development lending, supporting the “liberal credit policy”, which is also a view of Rajan (1994).
Comparing the results for state economic variables, state real GDP growth positively and significantly affects construction and land development, single-family (only in the case of the fixed-effects model) and multifamily residential and non-farmland and non-residential loans, with the income elasticities highest for construction and land development loans. Higher mortgage rates reduce single-family residential, non-farmland and farmland loans with the elasticity highest for the last category. In
JFEP 8,1
52
Table VII. Fixed-effects results
for different categories of real
estate loans
V ar
ia bl
es C
on st
ru ct
io n
an d
la nd
de ve
lo pm
en t
lo an
s Si
ng le
-f am
ily re
si de
nt ia
ll oa
ns M
ul ti
-f am
ily re
si de
nt ia
ll oa
ns M
od el
1 M
od el
2 M
od el
1 M
od el
2 M
od el
1 M
od el
2
C on
st an
t �
0 .2
3 6 **
(� 2.
12 3)
� 0 .2
6 1 **
(� 2.
26 2)
� 0 .3
4 4 **
* (�
3. 25
4) �
0 .2
9 9 **
* (�
2. 36
4) 0.
19 1
(1 .5
55 )
� 0.
00 9
(� 0.
06 6)
C ap
it al
-t o-
as se
ts �
0 .1
2 1 **
* (�
4. 07
6) �
0 .1
4 8 **
* (�
4. 33
8) �
0 .1
3 8 **
* (�
4. 70
7) �
0 .1
3 8 **
* (�
3. 66
) 0.
03 9
(1 .1
39 )
� 0.
04 2
(� 1.
01 9)
C re
di t
qu al
it y
� 0 .0
4 7 **
* (�
5. 33
1) �
0 .0
4 **
* (�
4. 72
) 0 .0
2 5 **
* (2
.9 48
) 0 .0
3 3 **
* (3
.4 41
) 0 .0
2 2 **
* (2
.2 15
) 0
.0 2
4 **
* (2
.3 34
) D
iv er
si fi
ca ti
on 0 .1
4 9 **
* (3
.0 97
) 0 .2
1 6 **
* (4
.6 42
) 0 .3
7 1 **
* (7
.7 1)
0 .4
0 **
* (7
.9 46
) 0 .2
0 3 **
* (3
.6 33
) 0
.2 4
7 **
* (4
.5 29
) O
C A
� 0 .2
4 2 **
* (�
3. 64
2) �
0 .2
6 6 **
* (�
4. 14
7) �
0 .8
8 7 **
* (�
13 .2
08 )
� 0 .9
2 3 **
* (�
13 .0
83 )
� 0 .6
1 2 **
* (�
7. 84
2) �
0 .6
1 **
* (�
7. 97
8) L
iq ui
di ty
� 0 .3
5 6 **
* (�
7. 91
3) �
0 .3
2 6 **
* (�
7. 53
4) �
0 .7
6 7 **
* (�
16 .5
52 )
� 0 .7
5 6 **
* (�
15 .6
13 )
� 0 .2
1 1 **
* (�
3. 92
4) �
0 .1
7 8
** *
(� 3.
40 )
R O
A 0 .0
6 5 **
* (4
.4 48
) 0.
01 2
(0 .8
06 )
� 0 .0
3 **
* (�
2. 03
5) �
0 .0
5 8 **
* (�
3. 50
6) �
0. 01
0 (�
0. 60
3) �
0 .0
3 2
** *
(� 1.
80 6)
R ea
lG D
P gr
ow th
1 .5
7 **
* (8
.3 72
) 0 .7
8 7 **
* (4
.0 98
) 0 .4
2 5 **
* (2
.5 33
) 0.
14 3
(0 .6
86 )
0 .4
9 8 **
* (2
.5 52
) �
0. 10
9 (�
0. 48
1) R
ea lm
or tg
ag e
ra te
s 0.
48 1
(1 .3
01 )
� 0 .6
4 **
* (�
1. 92
5) �
0. 53
9 (�
1. 39
6) In
fl at
io n
� 1 .1
7 9 **
* (�
4. 30
9) �
0. 44
8 (�
1. 47
4) �
1 .6
4 6
** *
(� 4.
99 9)
H P
I 1 .0
5 7 **
* (9
.6 39
) 0 .4
6 4 **
* (3
.8 61
) 0
.7 4
1 **
* (5
.6 86
) U
ne m
pl oy
m en
t ra
te s
� 0 .1
3 8 **
* (�
5. 34
3) �
0 .0
7 8 **
* (�
2. 77
5) �
0. 02
4 (�
0. 79
3) A
dj us
te d
R 2
0. 14
0 0.
20 6
0. 21
1 0.
21 9
0. 05
6 0.
08 3
F -s
ta ti
st ic
6. 11
3* **
8. 88
** *
10 .3
31 **
* 9.
75 2*
** 2.
02 4*
2. 72
5* *
D -W
1. 61
3 1.
73 3
1. 96
6 1.
99 3
2. 05
9 2.
01 2
N 1,
81 8
1, 81
8 2,
02 2
1, 86
9 2,
02 2
1, 86
9 (c
on ti
nu ed
)
53
Regional banking and
economic conditions
Table VII.
V ar
ia bl
es F
ar m
la nd
lo an
s N
on -F
ar m
la nd
no n-
re si
de nt
ia ll
oa ns
M od
el 1
M od
el 2
M od
el 1
M od
el 2
C on
st an
t 0 .4
9 3 **
* (3
.5 75
) 0 .3
0 3 *
(1 .8
8) 0.
06 5
(0 .9
16 )
0. 00
8 (0
.0 94
) C
ap it
al -t
o- as
se ts
0 .0
7 5 *
(1 .9
5) �
0. 04
7 (�
0. 97
8) �
0 .0
3 2 *
(� 1.
66 2)
� 0
.0 7
7 **
* (�
3. 09
2) C
re di
t qu
al it
y 0 .0
3 3 **
* (2
.9 23
) 0 .0
2 8 **
* (2
.2 89
) 0 .0
1 3 **
* (2
.2 62
) 0
.0 1
5 **
* (2
.4 96
) D
iv er
si fi
ca ti
on 0.
04 3
(0 .6
82 )
0. 04
9 (0
.7 58
) 0 .1
6 6 **
* (5
.1 71
) 0
.1 9
** **
** (5
.7 61
) O
C A
� 0.
13 9
(� 1.
58 5)
� 0.
10 8
(� 1.
20 5)
� 0 .2
9 **
* (�
6. 48
6) �
0 .3
0 1
** *
(� 6.
51 8)
L iq
ui di
ty �
0 .2
9 3 **
* (�
4. 83
3) �
0 .2
7 1
(� 4.
38 4)
� 0 .1
7 **
* (�
5. 51
6) �
0 .1
4 7
** *
(� 4.
64 2)
R O
A 0.
02 6
(1 .3
34 )
0. 01
5 (0
.7 28
) 0.
01 1
(1 .0
9) �
0. 01
1 (�
1. 04
2) R
ea lG
D P
gr ow
th �
0. 01
0 (�
0. 04
6) �
0 .5
1 4 **
* (�
1. 93
8) 0 .4
3 6 **
* (3
.9 02
) 0.
08 5
(0 .6
17 )
R ea
lm or
tg ag
e ra
te s
� 0.
45 8
(� 1.
05 8)
� 0 .8
4 0
(� 3.
79 5)
In fl
at io
n �
1 .7
5 6 **
* (�
4. 54
9) �
1 .0
5 4
** *
(� 5.
28 9)
H P
I 0.
14 4
(0 .9
35 )
0 .4
3 2
** *
(5 .4
74 )
U ne
m pl
oy m
en t
ra te
s �
0 .0
6 5 **
* (�
1. 81
2) �
0 .0
6 5
** *
(� 3.
50 8)
A dj
us te
d R
2 0.
02 7
0. 03
9 0.
08 2
0. 10
9 F
-s ta
ti st
ic 1.
92 8*
2. 21
8* *
3. 03
5* **
3. 67
** *
D -W
2. 14
9 2.
13 8
1. 90
3 1.
93 6
N 2,
00 8
1, 85
5 2,
02 2
1, 86
9
N o te
s: T
er m
s in
br ac
ke ts
de no
te t-
st at
is ti
c ba
se d
on ro
bu st
st an
da rd
er ro
rs cl
us te
re d
in st
at es
; *
, **
, **
* in
di ca
te s
si gn
ifi ca
nc e
at th
e 10
,5 an
d 1%
le ve
l; bo
ld va
lu es
de no
te th
e st
at is
ti ca
lly si
gn ifi
ca nt
co ef
fi ci
en tsJFEP
8,1
54
Table VIII. Dynamic GMM
estimation results for different categories of real estate loans
V ar
ia bl
es F
ar m
la nd
lo an
s N
on -F
ar m
la nd
no n-
re si
de nt
ia ll
oa ns
M od
el 1
M od
el 2
M od
el 1
M od
el 2
C on
st an
t 0
.5 5
9 **
* (2
.4 8)
0. 35
5 (1
.2 9)
0 .1
6 4
** (2
.0 2)
0. 06
4 (0
.8 0)
C ap
it al
-t o-
as se
ts 0
.1 3
6 *
(1 .8
3) �
0. 03
5 (�
0. 38
) 0.
02 5
(0 .9
5) �
0. 03
7 (�
1. 17
) C
re di
t qu
al it
y 0
.0 4
2 **
* (3
.5 3)
0 .0
3 3
** (2
.1 3)
0 .0
1 5
** (2
.0 1)
0 .0
1 5 *
(1 .7
7) D
iv er
si fi
ca ti
on 0.
02 0
(0 .2
3) 0.
03 0
(0 .3
6) 0
.1 9
5 **
* (2
.5 9)
0 .2
1 9 **
* (3
.0 9)
O C
A �
0 .3
6 4
** (�
2. 34
) �
0 .3
4 4
** (�
2. 37
) �
0 .3
3 4
** *
(� 3.
08 )
� 0 .3
3 7 **
* (�
2. 88
) L
iq ui
di ty
� 0
.3 2
5 *
(� 1.
65 )
� 0
.2 9
4 *
(� 1.
64 )
� 0.
11 4
(� 1.
06 )
� 0.
10 7
(� 0.
95 )
R O
A �
0. 00
7 (�
0. 30
) �
0. 00
6 (�
0. 27
) �
0. 00
2 (�
0. 15
) �
0 .0
1 9 *
(� 1.
79 )
R ea
lG D
P gr
ow th
� 0.
12 5
(� 0.
21 )
� 0.
84 8
(� 1.
22 )
0 .3
5 **
(2 .3
3) �
0. 04
1 (�
0. 29
) R
ea lm
or tg
ag e
ra te
s �
0 .9
8 5
** *
(� 3.
65 )
� 0
.9 2
1 **
* (�
5. 48
) In
fl at
io n
� 2
.5 1
2 **
* (�
6. 06
) �
1 .2
4 2 **
* (�
7. 87
) H
P I
0. 07
1 (0
.4 3)
0 .3
8 3 **
* (3
.3 8)
U ne
m pl
oy m
en t
ra te
s �
0 .0
8 9
** *
(� 3.
82 )
� 0 .0
5 6 **
* (�
2. 79
) y (
t– 1 )
� 0
.2 4
8 **
* (�
3. 24
) �
0 .2
5 8
** *
(� 3.
60 )
0. 02
9 (0
.4 7)
0. 01
1 (0
.1 9)
C hi
-s qu
ar e
16 5.
63 19
9. 68
78 .0
5 26
8. 01
N 1,
94 7
1, 79
4 1,
96 1
1, 80
8 A
R (1
) �
2. 80
6* **
� 2.
58 2*
** �
3. 98
7* **
� 3.
80 3*
** p-
va lu
es 0.
00 5
0. 00
98 0.
00 01
0. 00
01 A
R (2
) �
1. 27
4 �
1. 31
9 �
0. 80
6 �
1. 00
4 p-
va lu
es 0.
20 3
0. 18
7 0.
42 1
0. 31
5 ( co
nt in
ue d)
55
Regional banking and
economic conditions
Table VIII.
V ar
ia bl
es F
ar m
la nd
lo an
s N
on -F
ar m
la nd
no n-
re si
de nt
ia ll
oa ns
M od
el 1
M od
el 2
M od
el 1
M od
el 2
C on
st an
t 0
.5 5
9 **
* (2
.4 8)
0. 35
5 (1
.2 9)
0 .1
6 4
** (2
.0 2)
0. 06
4 (0
.8 0)
C ap
it al
-t o-
as se
ts 0
.1 3
6 *
(1 .8
3) �
0. 03
5 (�
0. 38
) 0.
02 5
(0 .9
5) �
0. 03
7 (�
1. 17
) C
re di
t qu
al it
y 0
.0 4
2 **
* (3
.5 3)
0 .0
3 3
** (2
.1 3)
0 .0
1 5
** (2
.0 1)
0 .0
1 5 *
(1 .7
7) D
iv er
si fi
ca ti
on 0.
02 0
(0 .2
3) 0.
03 0
(0 .3
6) 0
.1 9
5 **
* (2
.5 9)
0 .2
1 9 **
* (3
.0 9)
O C
A �
0 .3
6 4
** (�
2. 34
) �
0 .3
4 4
** (�
2. 37
) �
0 .3
3 4
** *
(� 3.
08 )
� 0 .3
3 7 **
* (�
2. 88
) L
iq ui
di ty
� 0
.3 2
5 *
(� 1.
65 )
� 0
.2 9
4 *
(� 1.
64 )
� 0.
11 4
(� 1.
06 )
� 0.
10 7
(� 0.
95 )
R O
A �
0. 00
7 (�
0. 30
) �
0. 00
6 (�
0. 27
) �
0. 00
2 (�
0. 15
) �
0 .0
1 9 *
(� 1.
79 )
R ea
lG D
P gr
ow th
� 0.
12 5
(� 0.
21 )
� 0.
84 8
(� 1.
22 )
0 .3
5 **
(2 .3
3) �
0. 04
1 (�
0. 29
) R
ea lm
or tg
ag e
ra te
s �
0 .9
8 5
** *
(� 3.
65 )
� 0
.9 2
1 **
* (�
5. 48
) In
fl at
io n
� 2
.5 1
2 **
* (�
6. 06
) �
1 .2
4 2 **
* (�
7. 87
) H
P I
0. 07
1 (0
.4 3)
0 .3
8 3 **
* (3
.3 8)
U ne
m pl
oy m
en t
ra te
s �
0 .0
8 9
** *
(� 3.
82 )
� 0 .0
5 6 **
* (�
2. 79
) y (
t– 1 )
� 0
.2 4
8 **
* (�
3. 24
) �
0 .2
5 8
** *
(� 3.
60 )
0. 02
9 (0
.4 7)
0. 01
1 (0
.1 9)
C hi
-s qu
ar e
16 5.
63 19
9. 68
78 .0
5 26
8. 01
N 1,
94 7
1, 79
4 1,
96 1
1, 80
8 A
R (1
) �
2. 80
6* **
� 2.
58 2*
** �
3. 98
7* **
� 3.
80 3*
** p-
va lu
es 0.
00 5
0. 00
98 0.
00 01
0. 00
01 A
R (2
) �
1. 27
4 �
1. 31
9 �
0. 80
6 �
1. 00
4 p-
va lu
es 0.
20 3
0. 18
7 0.
42 1
0. 31
5
N o te
s: T
er m
s in
br ac
ke ts
de no
te t-
st at
is ti
c ba
se d
on ro
bu st
st an
da rd
er ro
rs cl
us te
re d
in st
at es
; *
, **
, **
* in
di ca
te s
si gn
ifi ca
nc e
at th
e 10
,5 an
d 1%
le ve
l; bo
ld va
lu es
de no
te th
e st
at is
ti ca
lly si
gn ifi
ca nt
co ef
fi ci
en ts
JFEP 8,1
56
contrary, construction and land development loans are positively and significantly affected by interest rates. Higher inflation again reduces different categories of real estate lending, except for single-family residential loans where the effect is insignificant. Farmland loans exhibit the highest inflation elasticity. A rise in state housing prices significantly increases lending, except for farmland loans, with the effect most pronounced for construction and land development loans. Much like the results for aggregate real estate loan categories, higher state unemployment rates have a negative effect on disaggregated real estate loans, with the exception of multi-family residential loans, with the elasticities ranging between �0.06 and �0.14 per cent. Lastly, a rise in construction and land development loans in one year increases lending by 0.14-0.19 per cent for the next year. For farmland loans the lagged effect is negatively significant, while for the other loan types there is no evidence of persistence in lending[9].
5.3 Comparing real estate loans with other loan types Next, the sensitivity of other categories of commercial bank lending to both state-level banking and economic conditions are estimated. Such an analysis provides us with a comparative perspective with real estate loans. Table IX presents the results.
Unlike real estate loans, other loans types are insignificant to bank capitalization, except interbank lending that is negatively affected. Inferior credit quality reduces both agricultural and C&I loans, similar to construction and land development loans. Greater liquidity reduces individual loans, similar to overall real estate loans, but is insignificant in influencing other loan types. Greater diversification increases all categories of lending with the highest elasticity for interbank loans. Higher cost inefficiency significantly reduces individual, C&I and interbank lending, with the last category again been the most sensitive. Bank profits are statistically insignificant for all four loan types. The findings for diversification, liquidity, OCA and ROA are consistent with the findings for overall real estate loans.
Gleaning at the results for state-level economic variables, real GDP growth increases agricultural, C&I and individual loans, but reduces interbank lending. Likewise, higher inflation rates increase interbank lending. This suggests that during the downward phase of state business cycles, banks prefer to move toward the interbank lending market away from the other risky types of loans. Higher inflation and unemployment rates significantly reduce individual loans implying that during economic downturns, banks refrain in engaging auto, credit card or consumer loans, becoming more wary about their ability to repay such loans. The results for interest rates are contrary to those for overall real estate loans[10]. Both increases in real bank prime loan rate and real federal funds rate increase C&I and interbank lending, respectively. A rise in state-level housing prices does not influence these loans types, as one would expect, except for individual loans where a 1 per cent rise in state-HPI increases loans by 0.36 per cent. Lastly, higher agricultural and interbank lending in one year reduces lending in the next one. However, no significant lagged effects are found for other categories of loans.
6. Conclusion Debate about the behavior of bank lending has recently intensified in the popular press, as well as within academia, in the wake of the global financial crisis. More specifically, research on real estate loans have been at the forefront of analysis as banks excess loans portfolio toward the real estate sector has been identified as a key source of the financial
57
Regional banking and
economic conditions
Table IX. Dynamic GMM estimation results for other categories of loans
V ar
ia bl
es A
gr ic
ul tu
ra ll
oa ns
C &
I lo
an s
M od
el 1
M od
el 2
M od
el 1
M od
el 2
C on
st an
t �
0 .4
3 1
** (�
2. 26
) �
0. 41
7 (�
1. 55
) �
0. 16
1 (�
1. 23
) �
0. 18
9 (�
1. 05
) C
ap it
al -t
o- as
se ts
� 0.
04 7
(� 0.
93 )
� 0.
03 6
(� 0.
40 )
� 0.
00 2
(� 0.
03 )
� 0.
00 8
(� 0.
12 )
C re
di t
qu al
it y
� 0
.0 4
2 **
* (�
2. 59
) �
0 .0
4 1
* (�
1. 71
) �
0 .0
2 3
** *
(� 2.
54 )
� 0.
02 0
(� 1.
57 )
D iv
er si
fi ca
ti on
0 .1
6 8
** (2
.0 6)
0 .1
7 6
* (1
.7 3)
0 .2
4 4
** *
(3 .1
7) 0 .3
0 **
* (3
.7 8)
O C
A �
0 .5
3 6
** *
(� 3.
50 )
� 0
.5 1
4 **
* (�
2. 80
) �
0 .4
3 0
** (�
2. 27
) �
0 .4
8 2 **
(� 2.
19 )
L iq
ui di
ty 0.
01 9
(0 .1
3) 0.
04 2
(0 .2
7) �
0. 23
8 (�
1. 41
) �
0. 23
3 (�
1. 24
) R
O A
� 0.
01 4
(� 0.
81 )
� 0.
02 1
(� 1.
12 )
� 0.
00 8
(� 0.
59 )
� 0.
00 1
(� 0.
06 )
R ea
lG D
P gr
ow th
0 .3
8 8
* (1
.8 6)
0. 40
9 (1
.1 7)
0 .4
0 2
** *
(2 .7
5) 0.
00 3
(1 .5
2) R
ea li
nt er
es t
ra te
s 0.
70 3
(1 .3
4) 0
.6 3
1 **
* (2
.4 6)
In fl
at io
n 0.
08 8
(0 .1
3) �
0. 22
6 (�
0. 87
) H
P I
0. 26
6 (1
.0 6)
0. 23
4 (1
.5 7)
U ne
m pl
oy m
en t
ra te
s �
0. 01
9 (�
0. 34
) 0.
03 3
(0 .8
2) y (
t– 1 )
� 0
.0 8
8 *
(� 1.
97 )
� 0
.0 9
5 **
* (�
2. 66
) �
0. 00
3 (�
0. 12
) �
0. 00
6 (�
0. 24
) C
hi -s
qu ar
e 69
.8 40
52 .1
50 84
.3 80
78 .3
30 N
2, 18
5 1,
78 4
2, 21
5 1,
80 8
A R
(1 )
� 2.
96 1*
** �
3. 11
** *
� 3.
47 1*
** �
3. 43
6* **
p- va
lu es
0. 00
3 0.
00 2
0. 00
1 0.
00 1
A R
(2 )
0. 76
2 0.
32 1
� 0.
81 5
� 0.
83 9
p- va
lu es
0. 44
6 0.
74 8
0. 41
2 0.
39 7 (
co nt
in ue
d)JFEP 8,1
58
Table IX.
V ar
ia bl
es L
oa ns
to ot
he r
de po
si to
ry in
st it
ut io
ns In
di vi
du al
lo an
s M
od el
1 M
od el
2 M
od el
1 M
od el
2
C on
st an
t �
1 .2
9 8
** *
(� 3.
21 )
� 0.
41 7
(� 0.
82 )
� 0
.2 6
* (�
1. 79
) �
0. 31
8 (�
1. 64
) C
ap it
al -t
o- as
se ts
� 0
.5 5
5 (�
3. 76
) �
0 .4
7 0
(� 3.
16 )
� 0.
06 9
(� 1.
35 )
� 0 .1
1 3
(� 1.
70 )
C re
di t
qu al
it y
� 0.
03 5
(� 1.
06 )
0. 03
9 (0
.7 9)
0. 01
4 (0
.9 7)
0. 02
9 (1
.3 0)
D iv
er si
fi ca
ti on
0 .8
0 8
* (1
.9 2)
0 .9
8 8
** (2
.3 2)
0 .3
6 2
** *(
3. 05
) 0 .3
8 **
* (2
.8 2)
O C
A �
1 .4
0 1
** *
(� 3.
47 )
� 1
.6 0
4 **
* (�
3. 63
) �
0. 09
0 (�
0. 38
) �
0. 11
0 (�
0. 42
) L
iq ui
di ty
0. 37
6 (0
.7 4)
0. 45
6 (0
.9 2)
� 0
.6 8
** *
(� 3.
39 )
� 0 .6
9 9 **
* (�
3. 42
) R
O A
0. 07
0 (0
.8 3)
0. 11
6 (1
.3 2)
� 0
.0 2
7 (�
2. 05
) �
0. 01
3 (�
0. 76
) R
ea lG
D P
gr ow
th �
1 .8
9 7
** *
(� 3.
05 )
� 1
.9 2
2 **
(� 2.
33 )
0 .8
2 8
** *
(4 .1
1) 0 .4
4 1 *
(1 .6
4) R
ea li
nt er
es t
ra te
s 2
.7 7
5 **
(2 .1
1) �
0. 38
7 (�
1. 45
) In
fl at
io n
1 .5
0 3
* (1
.8 7)
� 0 .9
4 1 **
* (�
3. 31
) H
P I
� 0.
27 3
(� 0.
88 )
0 .3
5 7 **
* (2
.7 5)
U ne
m pl
oy m
en t
ra te
s �
0. 07
8 (�
0. 90
) �
0 .0
6 5 **
(� 2.
20 )
y ( t–
1 )
� 0
.1 7
1 **
* (�
4. 43
) �
0 .1
8 6
** *
(� 4.
76 )
0. 06
4 (1
.5 7)
0. 04
6 (0
.9 3)
C hi
-s qu
ar e
51 .1
50 87
.6 40
11 6.
83 0
14 9.
99 0
N 2,
06 5
1, 66
0 2,
21 5
1, 80
8 A
R (1
) �
5. 12
6* **
� 5.
03 8*
** �
3. 40
5* **
� 3.
48 7*
** p-
va lu
es 0.
00 0
0. 00
0 0.
00 1
0. 00
1 A
R (2
) �
0. 85
7 �
0. 33
3 �
0. 87
0 �
0. 92
0 p-
va lu
es 0.
39 0
0. 73
8 0.
38 5
0. 35
8
N o te
s: T
er m
s in
br ac
ke ts
de no
te t-
st at
is ti
c ba
se d
on ro
bu st
st an
da rd
er ro
rs cl
us te
re d
in st
at es
; *
, **
, **
* in
di ca
te s
si gn
ifi ca
nc e
at th
e 10
,5 an
d 1%
le ve
l; bo
ld va
lu es
de no
te th
e st
at is
ti ca
lly si
gn ifi
ca nt
co ef
fi ci
en ts
59
Regional banking and
economic conditions
crisis. The consequent disorderly deleveraging had an adverse impact on the ability of banks to support the real economy. The present study contributes to the real estate literature by examining the sensitivity of real estate loans given by commercial banks to state-level banking and economic conditions.
Summarizing the findings, banks capital, liquidity, overhead costs reduce real estate lending, while banks diversification and the size of the banking industry in each state increase such lending. Within disaggregated real estate loan types, construction and land development and single-family residential loans are most responsive to state banking and economic conditions. Moreover, real estate loans are found to be procyclical to state economic cycles with a rise in state real GDP growth, an increase in state HPI and a decline in both inflation and unemployment rates increasing real estate lending. Finally, comparing real estate loans to other categories of loans, the latter are found to be less sensitive to both state economic and banking conditions with interbank lending being countercyclical to state economic conditions.
The negative effect of capital on real estate lending imply as banks capital rises, it also increases their creditworthiness and franchise value, which incentivizes banks to engage in prudent lending and reduce real estate lending (including the most risky construction and land development, and non-farmland, non-residential loans). The finding that banks with lower capital-to-asset ratios (i.e. with higher debt-to-assets) are associated with relatively higher real estate loans, is consistent with the view that the excessive use of leverage by large bank holding companies is one of the factors that contributes to the development of the recent crisis in the USA. The findings for capitalization also bodes relevance when considered in the context of the recent discussions about capital adequacy ratios (Basel III). The adoption of a leverage ratio as an additional prudential tool to complement minimum capital adequacy requirements may well be able to reduce the risk of excessive real estate lending. In fact, the findings here support a recent push by the Federal Reserve to impose higher capital buffers on banks to mitigate risks. The positively significant coefficient of diversification is consistent with the views espoused in recent studies that higher earnings from non-interest income may propagate more risky lending practices.
Finally, regular stress tests of banks’ loan quality are increasingly based on macroeconomic assumptions to provide common scenarios for all financial institutions participating in such an exercise. Examining banks loan portfolio is a crucial part of such tests. Thus, the significant real estate loans elasticities with respect to credit quality imply a balance between the extent of portfolio diversification between risky real estate loans from other types, like consumer credit or investment securities is desirable. Bank stress tests of a general nature on the banking system in the USA (macro-tests) typically undertake a scenario analysis, where shocks or impact of specific variables on bank’s financial conditions are assessed. The statistically significant coefficients of state-level economic variables indicate that these underlying regional or “micro” economic determinants should be included when calibrating the impact of shocks on commercial banks real estate loans.
Notes 1. Agricultural loans are loans and advances made to finance agricultural production and other
loans to farmers. Commercial and Industrial loans represents those for commercial and
JFEP 8,1
60
industrial purposes to sole proprietorships, partnerships, corporations and other business enterprises. Loans to depository institutions represent loans to banks and other depository institutions, including domestic and foreign commercial banks, credit unions, mutual or stock savings banks, savings or building and loan associations. Loans to Individuals are all loans to individuals for household, family and other personal expenditures, including loans to finance autos, home improvement, medical expenses, personal taxes, household appliances, furniture, jewelry, education, student loans, mobile homes etc.
2. Construction and Land Development loans represents loans with maturities of 60 months or less made to finance land development or the on-site construction of industrial, commercial, residential or farm buildings. 1-4 Family Residential Properties: Represents permanent loans on 1-4 family dwelling units, mobile homes, individual condominiums and co-ops. Multifamily Residential loans are those on nonfarm properties with 5 or more dwelling units in apartments, housekeeping dwellings, co-operative type apartment buildings, and vacant lots in established multifamily residential sections. Farmland loans represent those secured by farmland, including improvements, and other land known to be used or usable for agricultural purposes. Non-farm non-residential loans are loans on business and industrial properties, hotels, motels, churches, hospitals, educational and charitable institutions, dormitories, clubs, lodges, association buildings, homes for aged persons, golf courses, recreational facilities and other similar properties.
3. Washington, DC is included in the sample to ensure the coverage of the largest possible number of banks under the Federal Reserve’s jurisdiction. However, other Federal Reserve covered territories like American Samoa, Guam etc. are excluded due to the unavailability of date on regional economic conditions.
4. Maddala and Wu (1999) argue that the individual unit root tests for panel data performs best when compared with tests that assume common unit roots, as it does not require a balanced panel data set. Hence, for purposes of robustness checks, I perform both common and individual panel unit root tests on the variables.
5. Hausman test provided a significant statistic for each and every specification, confirming the use of a fixed effects model over random effects.
6. When the state real GDP growth was replaced by state personal income growth, it had a positively significant effect on real estate loans as well, with the coefficient ranging between 0.55 and 0.89.
7. Peni et al. (2013) using the realized loan loss to measure risks find the average loss rate of 1.41 per cent on residential loans, 6.65 and 2.81 per cent, respectively, for construction and land development and commercial real estate loans, respectively, in year 2009.
8. Size of the banking industry was statistically significant under both estimation techniques. However, as noted previously, due to concerns of multicollinearity of industry size with the other banking-industry variables, the results presented are based on dropping the size variable.
9. AR (1) and AR (2) are the Arellano-Bond tests for first and second order autocorrelation of the residuals. One should reject the null hypothesis of no first order serial correlation and not reject the null hypothesis of no second order serial correlation of the residuals. In all GMM specifications, the requirements are met as suggested by the p-values of the AR (1) and AR (2) tests. These imply that the GMM results are consistent.
61
Regional banking and
economic conditions
10. For C&I loans I use the bank prime loan rate i.e. the rate commercial banks charge on a loan to a business, as the relevant interest rate while for interbank loans I use the Federal Funds rate.
References Arellano, M. and Bover, O. (1995), “Another look at the instrumental variable estimation of error
correction models”, Journal of Econometrics, Vol. 68 No. 1, pp. 29-51.
Barth, J., Caprio, G. and Levine, R. (2013), “Bank regulation and supervision in 180 countries from 1999 to 2011”, Journal of Financial Economic Policy, Vol. 5 No. 2, pp. 111-219.
Beck, R., Jakubik, P. and Piloui, A. (2015), “Key determinants of non-performing loans: new evidence from a global sample”, Open Economies Review, Vol. 26 No. 3, pp. 525-550.
Berger, A. and DeYoung, R. (1997), “Problem loans and cost efficiency in commercial banks”, Journal of Banking and Finance, Vol. 21 No. 6, pp. 849-870.
Berrospide, J.M. and Edge, R.M. (2010), “The effects of bank capital on lending: what do we know and what does it mean?”, Finance and Economics Discussion Series Working Paper 2010-44, Divisions of Research & Statistics and Monetary Affairs, Federal Reserve Board, Washington, DC, pp. 1-50.
Blundell, R. and Bond, S. (1998), “Initial conditions and moments restrictions in dynamic panel data models”, Journal of Econometrics, Vol. 87 No. 1, pp. 115-143.
Boudriga, A., Taktak, N. and Jellouli, S. (2009), “Banking supervision and nonperforming loans: a cross-country analysis”, Journal of Financial Economic Policy, Vol. 1 No. 4, pp. 286-318.
Buch, C., Eickmeier, S. and Prieto, E. (2014), “Macroeconomic factors and microlevel bank behavior”, Journal of Money, Credit and Banking, Vol. 46 No. 4, 715-751.
Contessi, S. and Francis, J.L. (2013), “US commercial bank lending through 2008: Q4 new evidence from gross credit flows”, Economic Inquiry, Vol. 51 No. 1, 428-444.
Dages, B., Goldberg, L. and Kinnery, D. (2000), “Foreign and domestic bank participation in emerging markets: lessons from Mexico and Argentina”, Federal Reserve Bank of New York Economic Policy Review, Vol. 6 No. 3, pp. 17-36.
Depken, C., Hollans, H. and Swidler, S. (2011), “Flips, flops and foreclosures: anatomy of a real estate bubble”, Journal of Financial Economic Policy, Vol. 3 No. 1, pp. 49-65.
Ghosh, A. (2015), “Banking-industry specific and regional economic determinants of non-performing loans: evidence from US States”, Journal of Financial Stability, Vol. 20, pp. 93-104.
Igan, D. and Pinheirp, M. (2010), “Exposure to real estate in bank portfolios”, Journal of Real Estate Research, Vol. 32 No. 1, pp. 47-74.
Im, K.S. Pesaran, M.H. and Shin, Y. (2003), “Testing for unit roots in heterogeneous panels”, Journal of Econometrics, Vol. 115 No. 1, pp. 53-74.
Ivashina, V. and Scharfstein, D. (2010), “Bank lending during the financial crisis of 2008”, Journal of Financial Economics, Vol. 97 No. 3, pp. 319-338.
Kashyap, A. and Stein, J. (2000), “What do a million observations on banks say about the transmission of monetary policy”, The American Economic Review, Vol. 90 No. 3, pp. 407-428.
Laeven, L. and Levine, R. (2009), “Bank governance, regulation and risk taking”, Journal of Financial Economics, Vol. 93 No. 2, pp. 259-275.
JFEP 8,1
62
Lee, K. and Lu, W. (2015), “Do bank regulation and supervision matter?: international evidence from the recent financial crisis”, Journal of Financial Economic Policy, Vol. 7 No. 3, pp. 275-288.
Levin, A., Lin, C. and Chu, C.J. (2002), “Unit root tests in panel data: asymptotic and finite sample properties”, Journal of Econometrics, Vol. 108 No. 1, pp. 1-24.
Louizis, D., Vouldis, A. and Metaxas, V. (2012), “Macroeconomic and bank-specific determinants on non-performing loans in Greece: a comparative study of mortgage, business and consumer loan portfolios”, Journal of Banking and Finance, Vol. 36 No. 4, pp. 1012-1027.
Lu, W. and Whidbee, D.A. (2013), “Bank structure and failure during the financial crisis”, Journal of Financial Economic Policy, Vol. 5 No. 3, pp. 281-299.
Maddala, G.S. and Wu, S. (1999), “A comparative study of unit root tests with panel data and new simple test”, Oxford Bulletin of Economics and Statistics, Vol. 61 No. 1, pp. 631-652.
Marcelo, A., Rodriguez, A. and Trucharte, C. (2008), “Stress tests and their contribution to financial stability”, Journal of Banking Regulation, Vol. 9 No. 2, 65-81.
Nkusu, M. (2011), “Nonperforming loans and macrofinancial vulnerabilities in advanced economies”, International Monetary Fund Working Paper No. 11/161, Washington, DC.
Peni, E., Smith, S. and Vahamaa, S. (2013), “Bank corporate governance and real estate lending during the financial crisis”, Journal of Real Estate Research, Vol. 35 No. 3, pp. 313-343.
Rajan, R. (1994), “Why bank policies fluctuate: a theory and some evidence”, Quarterly Journal of Economics, Vol. 109 No. 2, pp. 399-441.
Rioja, F., Rios-Avila, F. and Valev, N. (2014), “The persistent effect of banking crises on investment and the role of financial markets”, Journal of Financial Economic Policy, Vol. 6 No. 1, pp. 64-77.
Skorepa, M. and Seidler, J. (2015), “Capital buffers based on banks’ domestic systemic importance: selected issues”, Journal of Financial Economic Policy, Vol. 7 No. 3, pp. 207-220.
Stern, G. and Feldman, R. (2004), Too Big to Fail: The Hazards of Bank Bailouts, The Brookings Institution, Washington, DC.
Further reading Keeton, W. (1999), “Does faster loan growth lead to higher loan losses?”, Federal Reserve Bank of
Kansas City Economic Review, Vol. 84 No. 2, pp. 57-75. Keeton, W. and Morris, C. (1987), “Why do banks’ loan losses differ?”, Federal Reserve Bank of
Kansas City Economic Review, Vol. 72 No. 5, pp. 3-21.
Corresponding author Amit Ghosh can be contacted at: [email protected]
For instructions on how to order reprints of this article, please visit our website: www.emeraldgrouppublishing.com/licensing/reprints.htm Or contact us for further details: [email protected]
63
Regional banking and
economic conditions
Reproduced with permission of copyright owner. Further reproduction prohibited without permission.
- Do real estate loans reflect regional banking and economic conditions?
- 1. Introduction
- 2. Trends and patterns in commercial bank real estate lending in the USA
- 3. Determinants of bank lending in the US banking industry
- 4. Data description and preliminary statistical diagnostics
- 5. Econometric model, results and discussion
- 6. Conclusion
- References
