Tihana Škrinjarić
In my recent paper, I present a new model that helps assess risks in the UK housing market. Unlike traditional approaches that focus on average house price growth, the model estimates a full range of possible outcomes, allowing policymakers to identify potential risk of big house price drops. The analysis also highlights important regional differences: areas with more constrained housing supply tend to be more sensitive to changes in interest rates. Expanding supply can help ease price pressures. These insights can help improve the monitoring of housing market vulnerabilities and support financial stability policy.
What I do
Analysing vulnerabilities in the housing market is crucial to track financial stability risks stemming from this part of the economy. These developments are significant for banks, households, and firms, as housing represents a long-term investment. Furthermore, mortgages are one of the largest components of the balance sheets and cash flows of both UK lenders and households. However, forecasting dynamics in the housing market is challenging due to uncertainty around future outcomes. To do so, I collect many possible variables and indicators that could help me to predict house price growth from the perspectives of supply, demand, financial, and non-fundamental factors. I examine around 50 different indicators, which makes it the most comprehensive list in the house price modelling literature.
To gauge risks of big future house price drops, I apply a quantile regression approach (Koenker (2005)), and derive a measure of house-price-at-risk (HPaR) both at the national UK level and regional level. HPaR is a low percentile (I focus on the 5th) of the conditional distribution of future house price growth and therefore captures the severity of potential house price declines under adverse conditions. This approach allows me to examine how different parts of the distribution of house price growth – particularly the lower tail versus the median – are associated with changes in key variables, including:
- interest rates;
- income;
- debt burden dynamics;
- house price overvaluation, defined as significant increase in real estate prices beyond their intrinsic value, often driven by investor expectations that prices will continue to rise, even when fundamental factors cannot justify such valuations (Stiglitz (1990)). This overvaluation refers to house prices rising above levels that can be explained by economic fundamentals such as income, interest rates, credit conditions and housing supply. It therefore captures the part of house price growth that appears disconnected from these factors and is often associated with speculative expectations;
- supply constraints; and
- broader financial conditions.
In this way, the framework highlights that the relationship between these factors and house price growth can differ across normal and adverse states, without imposing a uniform effect across the distribution. The advantage of using quantile regression is that it more clearly captures periods of booms and busts compared to a standard linear regression model.
House price growth decomposition
Chart 1 presents the decomposition of the 5th percentile (I call it tail risk) nominal HPaR growth. The tail risk component effectively identifies downturns of early 1980s, 1990s, and dynamics of global financial crisis (GFC).
I observe that these declines were explained by different factors. In the early 1980s, the initial drop in nominal house prices was primarily linked to the oil price shock and a concurrent economic recession, followed by sharp increases in mortgage interest rates. The downturn in the early 1990s coincided with both a weakening economy and a subsequent correction in the housing market. While economic activity had already begun to slow before the housing bubble fully unwound, the decline in house prices likely amplified the recession through its effects on household balance sheets, consumption and credit conditions.
This is reflected by a sharp decline in house price overvaluation and compounded by a significant drop in credit activity and transactions in preceding quarters. During the GFC, transaction volumes explain most of the decline, followed by heightened financial stress and a contraction in credit supply. In the most recent downturn, the decline began with a slowdown in transactions, rising mortgage interest rates, and a drop in house price overvaluation. Both the predicted tail and median nominal house price growth have been trending downward since 2016.
Chart 1: Decomposition of nominal year-on-year (YoY) UK house price growth at the 5th percentile shows different contributions of house-price predictors across time
Note: Const – constant, Demand – includes YoY real GDP growth, Financial – includes YoY mortgage rate change, YoY change of price to income ratio, credit-to-GDP gap, CISS – composite indicator of systemic stress, YoY stock market growth, and YoY inflation; Non fund – house price overvaluation, Other – includes YoY house market transaction growth, CCI – consumer confidence index, and EPU – economic policy uncertainty; Supply – includes YoY housing investment growth, and YoY oil price growth.
Chart 1 also highlights a recurring pattern in which periods of elevated house price overvaluation are followed by subsequent corrections in tail house-price growth. This is consistent with the broader literature on asset-price cycles, which finds that prolonged periods of rapid price appreciation and overvaluation are often followed by market corrections as expectations adjust and prices converge back towards levels justified by fundamentals. In the decomposition, this mechanism is reflected in the non-fundamental component making a positive contribution during boom periods and a negative contribution during subsequent downturns. While house price overvaluation is not the sole driver of housing downturns, the results suggest that the unwinding of previous overvaluation amplified several of the observed declines in UK house prices.
Forward-looking measures of house price vulnerability
I calculate several forward-looking risk measures based on the estimated distributions for the one-year ahead model: distance to tail (measured as the difference between the median and tail risk forecasts), and the probability of negative growth – presented in Chart 2.
Distance to tail (left panel) measures the gap between the median and lower-tail forecasts. Larger values indicate a wider dispersion between central and adverse house price outcomes and are therefore often interpreted as a sign of increased vulnerability.
A notable spike in uncertainty is observed during the Covid-19 shock, though it dissipates quickly, as expected. The model is also successful in predicting such periods when house prices would drop significantly (right panel), as indicated by spikes of the forecasted probability series that preceded actual house price drops (grey shaded area).
Chart 2: Distance to tail and its decomposition (left), and probability of negative house price growth (right), one-year ahead
Notes: Left panel shows the difference between the median and tail growth (DTT = distance to tail). Right panel shows estimated probability of house price growth dropping below 0%. Grey shaded area denote periods when observed house price growth dropped below 0%. Estimates at a certain quarter of a year are based on information from the same quarter in the previous year.
What does regional analysis uncover?
Regional housing market vulnerabilities matter for financial stability because risks can build unevenly across the country and may not be fully captured by national indicators. The regional analysis shows that UK housing market vulnerabilities differ substantially across regions, highlighting the value of estimating separate HPaR models rather than relying solely on national results.
A key finding is that demand-related variables exhibit markedly different associations across regions. Income growth is most strongly associated with future house price growth in London, the South East, South West and East Anglia, suggesting that these regions are more sensitive to demand conditions than other parts of the UK (Chart 3, blue bars).
Chart 3: Differences between estimation results between regions
Notes: Bars denote the values of estimated parameters for selected variables, and lightly shaded blue, green, and orange bars denote statistically insignificant values.
Estimates of the relationship between mortgage rate changes and future house price growth vary considerably across the country. Supply-constrained regions, particularly southern and midland regions of the UK, display larger and faster coefficient of mortgage rate changes in the HPaR specification (Chart 3, orange bars). As a result, higher mortgage rates are associated with more pronounced risk of big house price drops in these areas.
On the supply side, greater housing supply is generally associated with lower future house price pressures in most regions (Chart 3, green bars). However, London, the South East and Scotland are exceptions. For the first two, the results align with the findings of Zahirovic-Herbert and Gibler (2014), who argue that in large, built-up metropolitan areas, new supply can lead to higher house prices. This is due to elevated land costs, stringent development constraints, and the potential need for brownfield remediation. Scotland has its own housing regulations and broader housing policy framework, which differ from those in England and Wales (Gibb (2019)).
This suggests that regional monitoring can provide valuable information for financial stability surveillance and policy assessment.
Key takeaways
While national estimates provide useful signals of housing market risk, regional results reveal some heterogeneity across regions. The associations between house prices and factors such as GDP growth, credit conditions and mortgage rates vary considerably across the UK, suggesting that both national and regional perspectives are useful for monitoring vulnerabilities. Stronger economic activity and housing supply are generally associated with lower downside risks, while higher mortgage rates and stronger credit growth are associated with greater risks of large house price falls.
Several limitations remain. Regional data availability is restricted, particularly for macrofinancial indicators, and the model is designed to identify predictive relationships rather than causal effects. Future research could incorporate richer regional data sets, explore housing market spillovers in greater detail, and investigate regional convergence clubs to better capture common housing market dynamics.
Tihana Škrinjarić works in the Bank’s Stress Testing and Resilience Division.
If you want to get in touch, please email us at bankunderground@bankofengland.co.uk or leave a comment below.
Comments will only appear once approved by a moderator, and are only published where a full name is supplied. Bank Underground is a blog for Bank of England staff to share views that challenge – or support – prevailing policy orthodoxies. The views expressed here are those of the authors, and are not necessarily those of the Bank of England, or its policy committees.