How interest rates affect spending: Household debt and the role of asset prices

Central banks adjust interest rates to affect consumption and manage aggregate demand. This column uses 6.8 million natural experiments created by differences in the timing of UK mortgage refinancing to study the impact of monetary policy on household spending behaviour. It finds that rate increases lead to lower cash-on-hand and lower household spending, whereas spending rises following rate cuts. These spending effects are driven by an asset price channel, whereby households borrow against higher house prices following rate cuts. Lower debt service costs matter less for explaining spending responses to interest rate changes.

The interest rate is one of the most powerful forces in the modern economy. Central banks wield it to manage aggregate demand. But how exactly do interest rates affect spending? Historically, economists first looked at the intertemporal trade-off channel: a higher interest rate raises the value of leaving a dollar in the bank to spend next year, relative to spending it today. However, as Hall (1988) influentially showed, the elasticity of intertemporal substitution is small, and subsequent work found little evidence that this channel is quantitatively important. Unsurprisingly, people do not shop more just because rates have dropped by half a percentage point.

So why do interest rates matter for spending, if at all?

A more promising angle is that interest rates operate less through the traditional intertemporal channel and more through financial conditions (Caballero and Simsek 2024, Caballero et al. 2024). Two aspects matter most: debt and asset prices. Household indebtedness is at historically high levels across every major economy. A rate change alters the debt service burden of indebted households, and hence their spending. Rate changes also move the asset prices that serve as collateral for borrowing, creating a second channel to spending. Both mechanisms may operate with long and variable delays, as different households experience rate changes at different times.

In a new paper (Foulis et al. 2026), we estimate how interest rate changes affect spending through these financial channels. But doing so requires overcoming a fundamental challenge: the world does not offer idealised experiments where some people, chosen at random, face a different interest rate change than others. As Eichenbaum et al. (2022) emphasise, the refinancing channel creates state-dependent effects of monetary policy. We exploit this by generating quasi-experimental variation using UK mortgages, which are typically fixed for two, three, or five years. ¹ When a deal expires, the mortgagor refinances at the prevailing rate. This creates a running natural experiment: every month, some mortgagors experience a rate reset while others do not, as illustrated in Figure 1.

Figure 1 Propensity of interest rate reset around mortgage refinancing month

There are 6.8 million such natural experiments, with mortgagors experiencing a rate change based on when their mortgage started versus prevailing rates, while most neighbours experience no change. Figure 2 shows the rich variation these shocks generate at the household level. By design, this variation identifies the effect of rate changes on households whose spending depends on whether and at what rate they can borrow. Our ten-year sample includes periods of both loosening and tightening. About three and a half million households experience a rate shock twice or more, enabling us to estimate effects within the same household.

Figure 2 Cross-sectional and time-series variation in household-level interest rate shocks

We link these natural experiments with administrative, monthly-level mortgage data from the FCA’s Product Sales Database and spending data from personal finance apps Money Dashboard and ClearScore (available for 70,000 refinancing deals). The left panel of Figure 3 displays the cash-on-hand impulse response to a one-percentage-point rate increase at refinancing. Cash-on-hand consists of the change in monthly payment (the passive component) and the change in total borrowing (the active component). A one-percentage-point rate rise causes cash-on-hand to fall by about £2,000 on average after one year. The right panel shows that household-level spending declines by £1500 (about 5% of consumption) over the same twelve-month period. If all deals expired at once, the implied spending change, without general equilibrium feedback, would be 0.7% of GDP after six months. This is broadly consistent with Cloyne et al. (2020), who show that mortgagors drive the aggregate consumption response to monetary policy.

Figure 3 Cash-on-hand and spending impulse responses to a one-percentage-point interest rate increase

These estimates imply that a decline in interest rates has a sizeable impact on spending via financial conditions. But how much of this effect reflects the direct burden of servicing existing debt, versus new borrowing encouraged by rising asset prices? Separating the asset price channel from the pure debt service channel matters. If interest rate cuts affect spending primarily by encouraging new debt against rising house prices, that raises a potential trade-off between stimulating demand and macroprudential concerns.

Separating the asset price channel poses a challenge: one needs independent variation in how sensitive asset values are to interest rates, i.e. ‘asset duration’. We exploit two sources of such variation. First, we use regional differences in local housing supply elasticity to instrument for housing duration. Second, we use variation in lease duration of properties in the UK to instrument for housing duration at the household level. Both approaches yield the same conclusion: spending rises mostly because households borrow against higher house prices; lower debt service after rate cuts matters less.

Figure 4 illustrates the power of the asset price channel. We separately estimate impulse responses for households in the lowest and highest decile of housing duration. House prices in the highest-duration regions are most sensitive to interest rates, while those in the lowest-duration regions are barely affected. The left panel shows the borrowing response to a one-percentage-point rate rise for each group; the right panel shows the spending response. The contrast is stark: households in low-duration regions are hardly impacted by the rate shock, while those in high-duration regions respond strongly. 

Figure 4 Borrowing and spending impulse response by housing duration decile

The asset price channel drives most of the spending effect, and most of that spending is financed by borrowing against rising home values. This is consistent with the ‘indebted demand’ channel of Mian et al. (2021). We estimate that the marginal propensity to consume and to borrow out of interest-rate-driven housing wealth are both about 0.04 for mortgaged households. This is in line with broader estimates of the marginal propensity to consume out of housing wealth of 0.03–0.05 (e.g. Campbell and Cocco 2007, Guren et al. 2021).

The role of debt service costs has often been emphasised in the transmission of monetary policy to consumption. Our results suggest that monetary policy affects consumption in large part through asset prices and borrowing. A second implication is that interest rate changes affect spending with ‘long and variable lags’ (Friedman 1961), as households gradually react to propagating interest rate shocks when they get an opportunity to refinance existing debts.

Source : VOXeu