Transcript How do demand determinants explain the credit growth in

Estimating Credit Demand in Croatia
By
Katja Gattin-Turkalj, Igor Ljubaj,
Ana Martinis, Marko Mrkalj
Discussant: K. Žigić
Prague, Czech Republic
Research Questions
How do demand determinants explain the credit growth in
Croatia?
Was credit growth excessive during the past 10 years?
Does recent boom in property prices help explain credit
demand in Croatia?
Importance: macroeconomic stability (inflation, foreign debt,
potential fragility of the banking sector)
Demand for total loans to the private sector versus households
demand for loans
Stylized Facts about Loans Developments

Three lending booms

Increased importance of leasing after HNB regulatory
measures
Modeling

Basic model:
Loans = f( r, GDP)

Extensions:
Loans = f( r, GDP, Loans_1, X)

OLS
Econometrics problems: (trend versus first-difference
stationary data, endogeneity/ causality issues, omitted variables,
multicolinearity, serial correlation)
Results
1. The baseline specification seems to explain well the
observed developments of credit;
2. The extension of the baseline equation, including the
variables used to control for the supply specific factors,
did not significantly change the results; in particular,
they did not change the assessment of periods of
“excessive” credit growth vs. periods when the credit
growth was in line with the fundamentals.
Results
3. As for household demand for credit, the results show
that the coefficient on the interest rates on household
loans was negative in all specifications (although nonsignificant in the specifications containing lagged
depending variable);
4. House price index did not contribute to explaining
household credit demand.
Comments, Suggestions, Questions
Why not try a simple structural supply-demand model for
loans?
(Only if banking sector can be characterized as
monopolistically competitive, estimation of demand is
not a problem!)
…or at least demand for different categories of loans
For instance, demand for leasing versus demand for an
ordinary loan (different monetary implications of these
two modes):
Loans = f( rlo, rle ,GDP)
Leasing = g( rle, rlo ,GDP)
Comments, Suggestions, Questions



Another promising possibility is to disaggregate
household demand into its components and look
deeper into its determinants (demographic factors, size,
state and age of the vehicle stock, etc. )
Estimate demand system for components of household
demand!
Based on Fig 4, (page 9) one would expect that wealth
has strong impact on the demand for households loans.
Comments, Suggestions, Questions


“ More formally, assumptions on exogeneity “determined outside
the system under analysis“ and causality in the Granger sense
"presence/absence of feedback between variables“ (Hendry,
1995) are made. Our expectation is that, although supply side
plays a significant role in financial sector transition, we can still
learn about the demand specific factors and their influence on
credit growth.” (Page 11)
What is the meaning of the section above? It seems
that there was no Granger causality testing but just
assumption of exogenaty among variables
Comments, Suggestions, Questions


“Supply effects have traditionally played more prominent role in
recent credit booms in the CEE and NMS, and therefore,
literature on "credit booms" and "lending channel" has
predominantly been focusing on the supply side factors.” (page 1)
How do you know that? References?
Therefore, although credit demand offers only partial explanation
of credit developments and hinges on the assumption on the
limited supply effects, an expanding strand of recent credit
literature focuses on the demand factors alone. (page 1)
(Again!) How do you know that? References?
Comments, Suggestions, Questions
Be more specific on how you measure the “excessiveness” of credit growth. It
seems, by checking if actual values exceed the ones predicted by the model. Is
this sufficient? Out of sample prediction has to be used here from some
other country!
What is the rationale for consumption to be used as regressor? (Perhaps, it
serves as a good proxy for household income!)
Define the notion of “GAP” used in the text (from table footnote, the reader
can learn that gaps are defined as the deviation from H-P trend. So content
of Fig. 5 and 9 have to be better explained!)
Finally, when using abbreviations like, for instance, ADF then indicate them in
the parentheses after the first use of it: e.g. Augmented Dickey-Fuller (ADF).
(Also H-P was nowhere defined, Hodrick-Presoctt filter, I presume)