Directions for future research

Throughout the discussion, we have highlighted a few gaps in the literature in the context of our framework. In this section, we summarize them. The best-measured object in the literature, the direct credit response of exposed households, is only a small part of the total response, and it does not reveal the full private burden of the policy, the equilibrium adjustments that follow, or the corrective benefit. In our view, one more estimate of this response is therefore less valuable than research on what happens to households that are kept out of the market, how much of the aggregate response comes from market adjustment, where displaced credit goes, and how large the external costs of household leverage actually are. We summarize this in the form of six areas for future research.

Measure the full, direct burden on households. Much of the direct credit response comes from the extensive margin, yet most of what we know about adjustment comes from households that still buy. A key question is therefore whether mortgage constraints delay homeownership by a year or two or exclude some households permanently. This distinction determines how the extensive margin should be interpreted in term (I). Linking mortgage applications and rejections to credit registers, housing transactions, rental contracts, and data on household and parental wealth would show how long entry is delayed, how much delayed households pay in rent, what and where they eventually buy, and whether family resources determine who enters (Wold et al. 2024; Benetton et al. 2022). Such data exist, for instance, in the Nordic countries. Followed over a longer horizon, the same data could also show whether tighter constraints change the age profile of homeownership and the transmission of housing wealth across generations. Since households that never buy do not reveal their willingness to pay for borrowing capacity through their choices, survey designs that elicit it directly, as in Fuster and Zafar (2021), are a natural complement.

Estimate direct and equilibrium effects in the same setting, and identify the channel. The aggregation gap in Section 5.2.2 is based on comparing studies of different countries, instruments, outcomes, and horizons, and on amplification estimates borrowed from other credit shocks. The tools needed to do better exist, but have rarely been applied to mortgage constraints. A study that estimates the household response within markets and the market-level response across markets with different average exposure, for the same outcome and horizon, delivers a regional multiplier directly (Mian et al. 2026; Huber 2023). Such studies should also report the diagnostics for heterogeneous spillovers suggested by Berg et al. (2021) and Huber (2023). To the best of our knowledge, no study of a borrower-based measure has properly estimated and analyzed the general equilibrium amplification of mortgage constraints. Mortgage constraints also offer a natural test of the mechanism. The collateral channel operates under LTV limits but not under income-based limits, so house-price spillovers to credit should be weaker for LTI-constrained households. Settings where both types of limits were introduced simultaneously can be used to test this. Loan-level data with lender identifiers would make it possible to separate propagation through housing markets from propagation through lenders (Sarto 2025b), and estimates of the segmentation between owner-occupied and rental markets outside the United States (Greenwald and Guren 2025) would help discipline the step from regional to national effects.

Trace where the displaced credit goes. A tightening only lowers financial vulnerability if the credit it removes does not reappear elsewhere in an equally fragile form. Section 5.1.3 discusses several examples of such leakage, including Irish banks expanding riskier corporate lending and mortgage lending in the United Kingdom (Acharya et al. 2022; McCann and O’Toole 2019). Households also find ways around the constraints, for instance by paying a large premium for unregulated debt (Karapetyan et al. 2024) or through parents borrowing against their own homes (Benetton et al. 2022). Most studies, however, measure lending in the regulated segment only. The relevant outcome for term (III) is the distribution of leverage and debt service across the whole system, including debt held by parents, by consumer-credit and non-bank lenders, and by foreign affiliates. Credit registers covering all lenders and household balance-sheet data that link generations make this measurable in several countries. Without such evidence, a fall in regulated mortgage credit is only weak evidence of a fall in aggregate vulnerability.

Study households when downturns occur and measure external costs of high leverage in a downturn. The main rationale for mortgage constraints is that high household leverage imposes costs on others. Several authorities explicitly refer to an aggregate demand externality, where highly indebted households cut consumption sharply after a shock and thereby reduce demand and incomes for everyone else (Section 2). The normative models behind this rationale are most useful for telling empirical researchers what to measure. In Farhi and Werning (2016) and Korinek and Simsek (2016), the externality is the product of how sharply indebted households cut spending after a shock and how far monetary policy fails to offset the resulting fall in demand. Neither object has been measured in a setting with mortgage constraints, and doing so requires two steps. The first is to establish whether households whose mortgages were originated under tighter constraints actually cut consumption less when house prices fall or incomes decline. Although evidence from outside the macroprudential literature supports this channel (Mian et al. 2013), the effect could differ for households constrained by regulation, since lower leverage comes together with lower liquid buffers (Van Bekkum et al. 2024; Aastveit et al. 2022). The second is to establish whether lower spending among indebted households reduces income and employment among other households. Regional evidence from the US housing bust and the Hungarian foreign-currency debt crisis suggests that such spillovers through local demand can be large (Mian and Sufi 2014; Verner and Gyöngyösi 2020), but we are not aware of studies that estimate them in a setting with mortgage constraints. The designs discussed above could be used for this purpose, by linking regulatory exposure at origination to outcomes in a later downturn for both exposed households and non-exposed households in the same local market. The second factor in the externality, the monetary offset, could be studied by comparing countries inside and outside a currency union, or periods with and without a binding lower bound on interest rates. The pecuniary externality in Dávila and Korinek (2018) calls for an analogous exercise: measuring how much forced sales by levered households move house prices, and how many other borrowers’ constraints tighten as a result, which is the collateral feedback in term (II).

Understand whether mortgage constraints keep countries out of the “R-zone.” Few tightenings occur in the joint credit and house-price boom where historical crisis risk is highest. As discussed in Section 6.1, this is consistent both with policy being used too late and with successful preemption. Separating the two requires an estimate of the path that credit, debt service, lending standards, and house prices would have followed without the tightening. The main obstacle is endogeneity, as authorities tighten in response to the very credit growth they cite as their objective. Policy series therefore need to be purged of their predictable component or identified narratively (Richter et al. 2019). With more than 150 tightening episodes across nearly 40 countries, event-study and synthetic-control designs on the crisis map are now feasible. The question such designs should answer is whether a tightening keeps an economy from entering the region where the link between vulnerability and crisis risk in term (III) is steepest.

Evaluate mortgage constraints jointly. Most studies evaluate one instrument at a time, but many countries use several at once. Norway combines LTV and debt-to-income limits with amortization requirements and an interest-rate stress test, Ireland introduced LTV and LTI limits together in 2015, and Sweden combines amortization requirements with an LTV limit. As discussed in Section 2, the instruments bind for different households, and the constraint that binds for a given household can change with house prices, income, and interest rates. The effect of tightening one instrument therefore depends on the others in place. A tighter LTV limit may, for instance, have little effect on a household that is already constrained by an income-based limit, but bind more often after house prices fall. The instruments also induce different adjustments, as LTV-constrained households draw on their savings while LTI-constrained households buy cheaper homes (Higgins 2026). Meta-analytic estimates suggest that the credit effect is larger when LTV and income-based limits are used together (Malovaná et al. 2025), but few studies show how the instruments interact for the same population of borrowers. Loan-level data that record all the relevant ratios at origination would show how many households are constrained by each limit, and by several limits at once, and how the response to a change in one limit depends on the others. Such evidence is needed to evaluate the overall stance of borrower-based policy rather than each instrument in isolation. A further extension would include lender based measures, such as bank capital requirements, as well, acknowledging that different types of macroprudential regulation target similar systemic vulnerabilities.