📈 economics

Construction-at-Risk: Financial conditions and downside risk in residential construction

This paper introduces a "construction-at-risk" framework demonstrating that broad financial conditions, particularly Chicago Fed measures, provide superior early-warning signals for downside risks in U.S. residential construction—specifically distinguishing between planning-stage permit declines and execution-stage start reductions—compared to traditional mortgage-rate benchmarks.

Original authors: Myeongwon Lee, Doojin Ryu

Published 2026-09-12

📖 6 min read🧠 Deep dive

CC BY 4.0

Original authors: Myeongwon Lee, Doojin Ryu

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Housing is more than just a place to live; it is a massive engine that drives the economy, connecting the money people borrow to the jobs builders hold and the materials factories produce. When this engine sputters, the effects ripple outward, affecting everything from local banks to national employment rates. For decades, economists have watched house prices to gauge the health of this engine, assuming that a drop in value signals trouble. However, prices are a final result, a balance point where what people want to buy meets what builders can supply. A stable price can sometimes hide a crumbling foundation, much like a calm ocean surface can conceal a storm brewing deep below. The real stress often begins not when prices fall, but when the pipeline of new homes slows down or when planned projects stall before they are even built. Understanding this early warning system requires looking at the construction process itself, which happens in distinct stages: first, when a project is approved on paper, and later, when the ground is actually broken and construction begins.

Researchers at Sungkyunkwan University have developed a new way to watch for these early signs of trouble, a method they call "construction-at-risk." Instead of waiting for house prices to crash or for the total amount of money spent on housing to drop, they focused on the two specific steps where a housing project can fail. The first step is the building permit, which represents the planning stage where a project enters the approved pipeline. The second is the housing start, which marks the moment that plan turns into physical construction activity. By treating these as separate measures, the team could distinguish between a slowdown in future plans and a breakdown in the actual execution of current projects. They asked a simple but critical question: can broad financial conditions, such as the overall health of credit markets and the availability of loans, predict a sharp decline in these construction numbers before the decline becomes obvious in the wider economy?

To answer this, the researchers built a forecasting system that looks specifically at the worst-case scenarios, or the "lower tail," of construction activity. Rather than trying to predict the average number of homes built, they modeled the conditions that lead to the lowest 20 percent of outcomes. They used data from the Chicago Fed, which tracks a wide range of financial factors including risk, credit availability, and leverage, to see if these broad indicators could warn of a coming slump. They compared this broad financial data against the more familiar tool of mortgage rates, which is what most people think of when they consider the cost of borrowing for a home. The study covered a vast period of time, analyzing monthly data from the early 1960s through 2026, allowing them to see how these relationships held up across different economic cycles, including major crises like the global financial crisis and the pandemic.

The results showed that broad financial conditions are indeed powerful early-warning signals, but only when looking at the risk of a downturn. When the researchers used the broad financial indices to forecast the bottom 20 percent of construction outcomes, they found a significant improvement in accuracy compared to using history alone. This improvement was most pronounced for housing starts, the execution stage, suggesting that when financial stress rises, it hits builders who are trying to start work harder than it hits those who are just planning. The data indicated that these gains were concentrated at the lower end of the distribution; the same financial indicators did not help predict the average or the best-case scenarios. This distinction is crucial because it confirms that financial stress acts as a specific filter for downside risk, not just a general driver of all housing activity.

A key finding of the study was that standard mortgage rates and interest spreads, while important, could not reproduce these early-warning signals. The broad financial measures contained information about risk and credit conditions that went far beyond the simple cost of a mortgage. This suggests that the danger to residential construction comes from a complex mix of factors, including the overall health of the banking system and the willingness of lenders to take on risk, rather than just the price of borrowing. The researchers found that their model was particularly effective during periods of high financial stress, where the ability to predict a sharp drop in construction activity improved even further. This implies that during turbulent times, the link between financial health and the ability to build homes becomes tighter and more visible.

The study also highlighted the difference between the planning stage and the execution stage. While broad financial conditions helped predict trouble in both, the signal was stronger for housing starts. This aligns with the reality that once a project has a permit, it still needs financing, labor, and materials to actually begin. If the financial system tightens, projects that are already approved may still fail to break ground. Conversely, a drop in permits signals that the future pipeline is drying up, which is a different kind of warning. By separating these two stages, the "construction-at-risk" framework offers a more nuanced tool for policymakers and lenders. It allows them to see if a slowdown is coming from a lack of future plans or a failure to execute current ones, providing a clearer picture of where the stress lies in the housing supply chain.

Ultimately, this research shifts the focus from the final price of a home to the health of the construction process itself. It demonstrates that the risk of a housing downturn can be measured and forecasted by watching the lower tail of construction quantities. The findings suggest that broad financial conditions are a vital component of this monitoring, offering a warning system that is more sensitive to the specific mechanics of building homes than traditional interest-rate benchmarks. For those who manage the economy or lend money, this approach provides a way to spot trouble in the residential supply process before it fully manifests in prices or employment numbers, offering a chance to respond while the problem is still in the pipeline.

Technical Summary: Construction-at-Risk: Financial Conditions and Downside Risk in Residential Construction

Problem Statement
Existing research on housing risk predominantly focuses on house prices, housing returns, mortgage credit, or aggregate residential investment. While these metrics are significant, they represent equilibrium outcomes shaped by both demand and supply, potentially masking stress on the production side of the housing market. House prices may remain stable even when the construction pipeline weakens or when approved projects fail to transition into actual construction. Furthermore, residential supply is produced through a staged, frictional process rather than immediate adjustment. Consequently, there is a lack of an out-of-sample (OOS) framework specifically designed to monitor the lower tail of residential construction quantities to distinguish between pipeline-stage weakness and execution-stage stress. This study addresses this gap by introducing a "construction-at-risk" framework.

Methodology
The study employs a recursive OOS quantile forecasting approach adapted from "growth-at-risk" literature to the housing supply process.

Data and Variables: The analysis utilizes monthly U.S. data for two stage-specific construction measures: PERMIT (building permits, representing the planning-stage pipeline) and HOUST (housing starts, representing the execution stage). Financial conditions are measured using the Chicago Fed National Financial Conditions Index (NFCI), the Adjusted NFCI (ANFCI), and official subindexes for Risk, Credit, and Leverage. Standard benchmarks include the 30-year fixed mortgage rate, 10-year Treasury yield, term spread, and mortgage spread.
Target Construction: To isolate the incremental predictive content of financial conditions, the authors remove the component of construction growth explained by the series' own past history. This is achieved by estimating an autoregressive (AR) model for monthly log growth rates and cumulating the resulting AR residuals over a forecast horizon ( $h$ ). The target variable is the cumulative AR residual, $y_{t+h}^{Y,h}$ .
Forecasting Model: The study estimates recursive conditional quantile forecasts for the 20th percentile ( $\tau=0.20$ ) of the cumulative target. The baseline forecast horizon is three months ( $h=3$ ).
Financial Predictors: The authors test individual financial indices and subindexes. Additionally, they construct recursive Risk-Credit-Leverage (RCL) Principal Components (PCs). PC1 captures the common movement across Risk, Credit, and Leverage, while PC2 captures the relative composition (contrasting Credit against Risk and Leverage). These components are estimated recursively to avoid look-ahead bias.
Evaluation: Performance is measured by the percentage reduction in OOS quantile loss ( $qR^2$ ) relative to a recursive historical lower-tail benchmark (the empirical $\tau$ -quantile of realized cumulative targets). Statistical significance is assessed using Newey-West HAC standard errors and moving-block bootstrap p-values, accounting for the serial correlation inherent in overlapping forecast horizons.

Key Results

Predictive Power of Broad Financial Conditions: Broad financial condition measures (NFCI, ANFCI, Risk, and Leverage) significantly reduce lower-tail forecast losses for both PERMIT and HOUST compared to the historical benchmark. The most robust performance is achieved using the RCL PC1+PC2 specification, which reduces loss by approximately 8.3% for PERMIT and 8.7% for HOUST.
Lower-Tail Concentration: The forecasting gains are concentrated at the 20th percentile. At the median (50th) and upper (75th) quantiles, the gains are negligible or negative. This indicates that financial conditions provide specific information about adverse construction outcomes rather than general movements in the distribution.
Distinction from Rate Benchmarks: Standard mortgage and interest-rate benchmarks (including the mortgage spread) fail to reproduce the lower-tail forecasting gains generated by broader financial-condition measures. The RCL specifications outperform the mortgage-spread benchmark by approximately 9–10% in loss reduction, suggesting that construction downside risk contains information beyond simple borrowing costs or interest-rate channels.
Stage-Specific Insights: While both measures benefit from financial condition data, the gains are generally stronger for HOUST (execution stage) than for PERMIT (planning stage), particularly during periods of high financial stress. This suggests that financial conditions are especially informative when planned projects transition to actual construction, a stage sensitive to financing availability and balance-sheet constraints.
Robustness: The results hold across different forecast horizons (1 to 12 months), when predictors are lagged by one month to ensure information availability, and when major crisis periods (Global Financial Crisis and COVID-19) are excluded from the evaluation sample.

Significance and Contributions
The paper claims four primary contributions to housing and macro-financial research:

Stage-Specific Risk Measurement: It introduces a framework that distinguishes between pipeline-stage risk (permits) and execution-stage risk (starts), acknowledging that housing supply is a dynamic, frictional process.
Lower-Tail Focus: It demonstrates that the predictive content of financial conditions for construction is concentrated in the lower tail of the distribution, supporting a "construction-at-risk" interpretation where financial conditions signal adverse outcomes rather than average activity.
Beyond Rate Channels: It establishes that broad financial conditions (specifically the joint information in Risk, Credit, and Leverage) contain incremental predictive content for construction downside risk that is not captured by standard mortgage rates or interest-rate spreads.
Practical Monitoring Tool: It provides a transparent, implementable early-warning tool for policymakers and market monitors. By tracking lower-tail forecasts of permits and starts separately, stakeholders can identify deterioration in the future supply pipeline or stress in current construction execution before these issues are fully reflected in house prices or aggregate investment figures.

The study concludes that residential construction downside risk is a distinct and measurable housing-supply risk, and that macro-financial conditions are particularly useful for identifying the lower-tail deterioration of this supply process when viewed through a staged production lens.

Technical Summary: Construction-at-Risk: Financial Conditions and Downside Risk in Residential Construction

More like this