Applied Data Science Lab · MLW-BW · University of Hohenheim ·
Supervisor: Prof. Aderonke Osikominu
Summary
We build a monthly Housing Finance Tightness Index (HFTI) for Germany
from Bundesbank MFI statistics and measure how lending responds in the
next few months when tightness changes. Inputs are rate level, monthly
rate change, and loan volume growth. We cross-check with PCA and
estimate a small distributed-lag model for short-run effects.
Data
Source: Deutsche Bundesbank, MFI interest rate
statistics (BBIM1), new business to private households, monthly.
Main series: mortgage rate with initial fixation
>10y (level), total new housing loan volume.
Units: rates in % p.a. and volumes in million EUR.
Index construction
Inputs (monthly):
-
\(r_t\): mortgage rate (initial
fixation > 10y)
-
\(\Delta r_t = r_t - r_{t-1}\):
one-month rate change
-
\(g_t = \Delta \log V_t = \log V_t - \log
V_{t-1}\): loan-volume growth
Standardisation: for any series \(x_t\), define \(\tilde{x}_t = \dfrac{x_t -
\mu_x}{\sigma_x}\).
Index:
\[ \mathrm{HFTI}_t =
\frac{1}{3}\,\Big(\tilde{r}_t + \widetilde{\Delta r}_t -
\tilde{g}_t\Big). \]
Higher values indicate tighter housing finance conditions.
Short-run response
Outcome: \(y_t = \Delta \log
V_t\) (monthly loan-volume growth).
Model (distributed lags):
\[ y_t = \alpha +
\beta_{0}\,\mathrm{HFTI}_{t} + \beta_{1}\,\mathrm{HFTI}_{t-1} +
\beta_{2}\,\mathrm{HFTI}_{t-2} + \beta_{3}\,\mathrm{HFTI}_{t-3} +
\varepsilon_t . \]
Key effects:
-
On-impact: \(\beta_0\)
(same-month response).
-
Cumulative (1–3 months): \(\beta_1 + \beta_2 + \beta_3\).
Estimation by OLS; we report point estimates with 95% confidence
intervals.
PCA check
We also build a PCA index (PC1) from the same inputs and compare it to
HFTI to confirm both tell the same story.
Takeaways
-
HFTI tracks tightening and loosening and aligns with PCA PC1.
-
Tighter conditions reduce lending growth within a few months.
-
Effects can be stronger when rates are already high.
