Introduction
Forecasts of real estate collateral valuations represent a key parameter of Loss Given Default (hereinafter referred to as “LGD”) models for secured portfolios under IFRS 9 and as such expected credit loss (ECL) estimates are sensitive to fluctuations of real estate prices.
In this context, Residential Real Estate (RRE) collaterals are generally characterized as more homogeneous although geographical location can be a significant differentiator of price while Commercial Real Estate (CRE) collaterals are characterized with higher degrees of heterogeneity stemming primarily by numerous determinants such as geographical location, type (i.e. office, industrial real – estate, hotel, warehouse e.tc.) and quality.
The COVID-19 outbreak and its potential impact on real estate prices have raised concerns about the implications of fluctuations in collateral valuations in the ECL calculation.
With regards to the Commercial Real Estate (CRE) market, the following key trends are expected:
- Demand for office space may initially drop as remote working is expected to become a more permanent feature in the business world, however this could be counterbalanced following a redesigning of office premises with less desk-space per full time employee (FTE);
- Shift from retail to e-commerce may lead to store closures and consequently to an increased demand for logistics space;
- Tenant-related problems (in offices, malls, etc.) may lead to lease renegotiations, while owner-occupiers may strive to sell and lease back their assets or opt for synergies from properties/activities optimization in order to generate sufficient liquidity.
Based on the above the reliability and level of granularity of real estate forward looking indexation as a means of calculating more accurately ECL and capturing portfolio idiosyncratic characteristics in an attempt to optimize results remains highly topical.
Background information
In accordance with para 5.5.17, IFRS 9 requires Expected Credit Losses (ECL) to be measured in a way that reflects an unbiased and probability weighted amount by evaluating a range of possible outcomes incorporating all reasonable and supportable information.
For secured lending exposures, LGD estimates should take into account as a bare minimum the forecasts of future collateral valuations and the time to realization of collateral (GPPC 2.5.2.3).
Based on the above, fluctuations on real estate collateral values may have dramatic implications to ECL impairment allowance estimates. In this context, the existence of reliable and granular price indicators is of paramount importance to the enhancement of the accuracy of forward looking estimations.
Limitations stemming from the use of single aggregated indices
The common practice observed in the Greek market is the use of single aggregated indices (one for RRE and one for CRE) for the purposes of forward – looking valuation forecasts in the context of IFRS 9 ECL measurement. A property price index measures the changes of real estate prices as a percentage change from some specific reference date (on which index equals to 100). Those indices can be useful to predict the estimated market value for individual properties for different time horizons or to calculate the current market value for properties that have been valued at an older valuation date.
LGD estimates for secured portfolios, which incorporate forward – looking real estate valuations, are subject to model errors due to aggregation bias i.e. the situation where different types of properties from different locations, sub-sectors and quality types are aggregated into one single pool and this aggregation tends to obscure significant differences in value.
As a result, the suitability of an index (either for RRE or CRE assets) depends on its predictive accuracy as well as the extent and distribution of the estimation error which is highly dependent on aggregation bias.
Based on the analysis above it can be inferred that, CRE assets can be more susceptible to aggregation bias and thus more granular approaches should be considered in order to reduce model errors. However, depending on the portfolio and collateral characteristics (e.g. geographical concentration) a more granular approach could also be more efficient in the case of RRE assets especially for the purposes of IFRS 9 ECL calculation where estimation unbiasedness is an essential element.
Real estate assets: empirical evidence, relevant analysis & Indications
Real estate price indices are positively correlated with the evolution of the Greek economy as shown in the diagram below. Gross Domestic Product (GDP) growth stood at 1.4% in 2017 in Greece after 9 years of recession , while residential real estate prices increased by 1.7% in 2018 after 9 years of decline.
The Retail Price Index illustrates the greatest sensitivity to GDP movements (with a correlation of 98%), followed by the Residential Price Index (with a correlation of 92%).
The above suggest an expected decline in retail prices due to the negative implications of the COVID-19 pandemic with respect to Greek GDP.
In general, it can be expected that the majority of granular indices will present statistically significant divergence from the equivalent aggregated index, based mainly on the following three factors:
- Geographical location: It is considered as a primary driver of divergence among real estate assets (both RRE and CRE) and empirical evidence suggest that some locations such as urban areas or islands which are considered as tourist attractions substantially outperform the equivalent aggregate index, whilst some other underperform.
- Real Estate sub-type: Specifically, with regards to CRE assets, sub-sector is of paramount importance given the high degrees of heterogeneity observed i.e. hotels, offices, industrial real estate, and warehouses.
- Quality: Quality can vary across locations and real estate sub-types and is determined by the age and the fundamental characteristics of the property.
In this context, we analyzed a sample of 21 granular indices that represent Residential Price Indices of specific locations within the residential real estate sector of Athens, utilizing an annual dataset from 2010 until 2017. We compared the variation of each index with that of the aggregate index (Residential Price Index in Athens – as published by BoG) and observed that the majority of those granular indices present statistically significant divergence from that aggregated index.
The diagrams below display the results of our analysis in graphical form. The columns represent the percentage difference of the parameter estimate or ‘beta’ of a particular index compared to the aggregate index. The green columns show the difference when the particular index is experiencing a less sharp decline (for the period 2010-2017) compared to the aggregate index and the grey columns show the difference when the decline is more severe during the same period .
For instance, the price index in the southern suburbs of Athens, presents a cumulative decline during 2010-2017 of 17% less than the aggregate index, while on the contrary the price index in the western suburbs of Athens underperformed the aggregate index by 9%.
The discrepancy is more austere in CRE prices, since location, specifications and credibility of tenant are fundamental drivers for property prices. For example, office prices in Marousi, Athens have declined less than the aggregate Office Price Index, since demand has remained robust during the recent economic recession in Greece. The same applies for prime retail assets on the most prestigious high streets in the center of Athens (Ermou, Voukourestiou, Metaxa streets, etc.).
In this context, we constructed a series of indices based on the location of a sample of 80 actual transactions of office/mixed-use office assets that took place in a five-year period, from 2015 until 2019 in Athens. We compared the variation of each index representing different locations with that of the aggregate index (Office Price Index – as published by BoG). We observed that the granular indices present statistically significant divergence from the aggregate index.
For instance, the office price indices in the Marousi area and the Center of Athens, present a cumulative increase during 2015-2019 of 38% and 49% respectively, higher than the aggregate index, while on the contrary the price index generally in Attica underperformed the aggregate index by 3%.
The same applies for hotels, since Average Daily Rate (ADR) levels and valuations fluctuate among locations and do not follow a homogeneous downward or upward trend. The diagram bellow indicates the difference in Market Values per room for different hotel clusters (based on geographical criteria) in a sample portfolio of more than 150 hotel properties in Greece, with their corresponding average.
Conclusion
Based on the empirical evidence illustrated in the analysis above, it can be inferred that use of a single aggregated property index might not be appropriate in order to capture asset specific idiosyncratic characteristics stemming from divergence in location, sub-type of property, and asset quality and thus it can result to material bias.
Furthermore, more granular approaches to forward - looking indexation, could enhance accuracy while ensuring compliance with IFRS 9 requirement which is to take into consideration “all reasonable and supportable information” for the measurement of ECL.
Finally, it is worth noting that in some examples of our analysis, the granular approaches followed, on average, can result to ECL optimization, compared to the use of a single aggregate index thus reducing LGD model errors.
Get in touch
