1. ASSESSING FINANCIAL STABILITY RISKS ARISING FROM THE REAL ESTATE MARKET IN ITALY
Wanda Cornacchia | Banca d’Italia – Financial Stability Directorate
SAPIENZA UNIVERSITÀ DI ROMA
26 NOVEMBRE 2016
CONVEGNO SCIENTIFICO
LA SOCIETÀ ITALIANA E LE GRANDI CRISI ECONOMICHE

2. DISCLAIMER
The views expressed in this presentation are my own and do not necessarily coincide with those of Banca d’Italia.

3. OUTLINE The real estate market and financial stability in Italy
Risk assessment framework
early warning models (EWM) for Italy
real-estate related indicators
decision making based on guided discretion
Conclusions

4. The real estate market and financial stability in Italy
The real estate sector plays a central role in the Italian economy, due to (chart):
the contribution that it makes to production
the preponderance of real estate assets in households’ wealth
the links with the financial system
A significant share of bank lending is directed to this sector. (figure)

5. Risk assessment framework: data description (lhs variable)
Italy has not experienced any real estate-related systemic banking crises (i.e. non-crisis country) => major challenge is the definition of an appropriate “left-hand side variable”
In order to identify systemic banking vulnerabilities stemming from the real estate sector we therefore construct the following ratio:
𝑎𝑛𝑛𝑢𝑎𝑙 𝑓𝑙𝑜𝑤 𝑜𝑓 𝑛𝑒𝑤 𝑏𝑎𝑑 𝑑𝑒𝑏𝑡𝑠 𝑐𝑎𝑝𝑖𝑡𝑎𝑙 𝑎𝑛𝑑 𝑟𝑒𝑠𝑒𝑟𝑣𝑒𝑠
The ratio is available since 1990Q1 for households (i.e. the residential real estate sector, RRE) and
constructions and real estate agencies (C&RE).

6. RRE and C&RE banking vulnerability indicator (quarterly data; per cent)

7. Risk assessment framework: data description (potential EWIs)

8. Risk assessment framework: early warning models (EWMs)
3 complementary EWMs have been considered:
binary logit model
ordered logit model
Bayesian Model Averaging (BMA) based on linear regression models
Logit models provides an interesting and potentially useful tool to understand potential vulnerabilities in the RRE market but there are some drawbacks:
in order to define vulnerabilities we discretize a continous variable -> not all the information set is considered
the analysis is made only in sample (too few observations)
We focus on continuous left hand side variables and we apply BMA models to our analysis
BMA can be used for ranking variables: what are the most important variables?
easy to extend to out-of-sample analysis

9. Risk assessment framework: BMA
BMA tackles the problem of model and variable selection by estimating models for all possible combinations of variables and constructing a weighted average over all the models. If there are K potential variables, 2K variable combinations are available => 2K models are estimated.
By using the BMA, it is possible to associate each variable with a posterior inclusion probability (i.e. the sum of posterior model probabilities for all the models where a variable was included) => ranking of the indicators in terms of their importance in the BMA estimation.
Our methodology is based on 3 steps:
the selection of the optimal set of variables: feature selection
the variables that are correlated with our vulnerability indicator and not highly correlated with each other in the training period are selected.
starting from these variables, we estimate BMA linear regression model on the training period (1990Q1-2005Q4) and we keep the subset of variables that minimize the average prediction error, expressed in terms of the root mean squared error, for the test period (2007Q1-2014Q2) => optimal set of variables. (figure RRE; figure CRE)

10. Risk assessment framework: BMA
the estimation and out-of-sample performance evaluation
estimate the BMA linear regression model on the training period using the optimal subset of variables selected in step 1 and apply the recursive approach to evaluate the out-of-sample performance of the model. (figure RRE; figure CRE)
an out-of-sample forecasting exercise
Use the BMA model estimated on the whole observed period to forecast the average value together with the percentiles of the distribution of the vulnerability indicator. (figure)

11. Risk assessment framework: real estate related indicators
Main real estate related indicators to monitor financial stability risks in Italy:
Real estate indicators
Bank and credit indicators
Household indicators
Share of municipalities according to
house price variations (old)
(biannual data; percentage values)
Commercial property prices in Italy
(annual data; indices 2007=100)

12. Loans to construction and real estate agencies
(per cent)

13. New bad loans to households for house purchase by year of disbursment
(millions of euros and per cent)
New bad loans to households for house purchase by year of disbursment
(per cent of the number of contracts)

14. Conclusions
In Italy bank vulnerabilities related to real estate market downturns mostly originate from the real estate business sector, characterized by high debt, and to a much lesser extent from the household sector, whose indebtedness level is limited, also compared with other countries.
In the first half of 2016 the flow of new bad debts continued to decline for construction and real estate firms and remained low for home mortgages, including in relation to banks’ capital.
Early warning indicators point to a significant fall over the next few quarters in the risks for banks attributable to the real estate sector as regards loans both to households and to firms.
The econometric analysis is complemented with a set of real-estate related indicators developed to monitor financial stability risks. The results provided by these indicators are, among others, that: i) the share of Italian municipalities that recorded a decline in house prices on a cyclical basis decreased; ii) the riskiness of more recent borrowers is far below that of the borrowers that took out a loan before 2009.

15. References
Cornacchia W., Ferrari S. and Pirovano M. (2015), “Identifying early warning indicators for real estate banking crises”, ESRB Occasional Paper n.8.
Ciocchetta F., Cornacchia W., Felici R. and Loberto M. (2016), “Assessing financial stability risks arising from the real estate market in Italy”, Bank of Italy Occasional Paper n.323.
ESRB report on RRE and financial stability in the EU (2015), December.
ESRB report on CRE and financial stability in the EU (2015), December.

16. ANNEX

17. Nexus between the housing market and the economy
Source: ESRB report on RRE and financial stability (december 2015)
back

18. back

19. back

20. back

21. back

22. back

23. back

24. Share of municipalities according to house price variations (biannual data; percentage values)
back