Estimating Credit Demand in Croatia

1
Estimating Credit Demand in Croatia

• Katja Gattin-Turkalj, Igor Ljubaj,
• Ana Martinis, Marko Mrkalj

2
Overview of the Presentation

• Credit developments in Croatia
• The econometric model
• Total loans to the private sector
• Household loans
• Concluding remarks

3
Motivation
Transition of the financial sector towards the
end of 1990s predominantly supply side effects
demand factors gained greater importance

• GDP growth increases demand for loans via income
  effect and via wealth effect
• improving overall economic conditions, growing
  optimism by consumers and enterprises and sharp
  decline of interest rates, real convergence

4
Three phases of banks loan developments and the
CNB responses
5
Decline of interest rates contributed to credit
growth
6
Loans to households driven by housing loans
7
Increasing role of housing loans highlights the
risks associated with house price developments
8

• 2. Loans to the private sector
• - the model

9
Loans and GDP (gaps) exhibit a high degree of
correlation

10
ADF and PP tests
11
The model baseline equation
-0.06 3.09 -8.75 (-1.29)
(5.05) (-3.52)
12
GDP coefficient is in line with other studies

• Baseline result of GDP elasticity of 3.09
• Calza et al. (2001) 1.34 (eurozone)
• Calza et al. (2003) 1.60 (eurozone)
• Hofmann (2001) between 1.04 and 2.49 (16
  industrial countries)
• Hülsewig et al. (2001) 1.11 (Germany)
• Brzoza-Brzezina (2005) between 1.45 and 3.39
  (six European countries)

13
IR coefficients seem to vary more

• Baseline result of real interest rate elasticity
  of -8.75
• Calza et al. (2001) -1.01 (eurozone)
• Calza et al. (2003) -5.05 (eurozone)
• Hofmann (2001) between -0.01 and -0.08(16
  industrial countries)
• Hülsewig et al. (2001) -0.69 (Germany)
• Brzoza-Brzezina (2005) between -4.42 and -10.81
  (six European countries)

14
Bootstrapping method
15
Bootstrapping results for ß1
16
Bootstrapping results for ß2
17
Extensions of baseline specification
18
Standardized residuals of equations 1 through 12
19
Standardized residuals of equations 1 through 4
20
Extensions of baseline spec. Eq. 1 to 6
21
Standard specification tests

• Autocorrelation introducing persistence through
  lagged dependent variable, improves the fit,
  although residuals remain "noisy"
• Variance inflation factor1.3 ?logYt ß0
  ß1?irt ei 1/(1 R2)
• Test for omitted variables was done for all
  variables other than baseline regressors to test
  the explanatory power of the additional
• loanst-1 (Ho not omitted rejected at 1), the
  trend variable z (Ho rejected at 5), CPI (Ho
  rejected at 10 significance), and exchange rate
  (Ho not rejected)
• AIC and Schwartz criterion, systematically favor
  more parsimonious specifications

22
Recursive coefficients of equation loans -0.06
3.09gdp(-8.75)ir
23
Results

• the baseline specification seem to satisfactory
  explain the observed developments of credit
• the extension of the baseline equation, did not
  significantly change the results
• credit growth during the lending boom of the late
  1990's and in 2006 remains above the fitted line
  in all specifications and during the recession
  that followed the first lending boom credit
  plunged well below the fitted line
• inclusion of the lagged dependent variable
  "smoothes" the curve and improves the fit, but
  even then actual growth remains slightly above
  the fitted line until 1998Q1 and in 2006.

24
ResultsLoans (yearly growth rates) - baseline
specification
25

• 3. Household loans
• - the model

26
Correlation between household loans and
explanatory variables
27
Wages seem to move in the opposite direction from
loans

• Wage and households loans gaps (normalized)

28
as well as house price index - possibly due to
low data frequency

• YoY growth rates of house prices and household
  loans (normalized)

29
Equations for household loans
30
Concluding remarks

• the behavior of loans can be explained mainly by
  the developments of real GDP and real interest
  rates
• GDP captures most important forces behind the
  loan demand
• somewhat unexpectedly, house price index did not
  contribute to explaining household credit demand

31
Q?