How Do Mortgage Rate Resets Affect Consumer Spending and Debt Repayment? Evidence from Canadian Consumers – Research Dept. Working Paper No. 2206

Working Paper 2206 May 2022

Research Department

https://doi.org/10.24149/wp2206

Working papers from the Federal Reserve Bank of Dallas are preliminary drafts circulated for professional comment.

The views in this paper are those of the authors and do not necessarily reflect the views of the Federal Reserve Bank

of Dallas or the Federal Reserve System. Any errors or omissions are the responsibility of the authors.

How Do Mortgage Rate Resets

Affect Consumer Spending and

Debt Repayment? Evidence from

Canadian Consumers

Katya Kartashova and Xiaoqing Zhou

How Do Mortgage Rate Resets Affect Consumer Spending and

Debt Repayment? Evidence from Canadian Consumers

Katya Kartashova

†

and Xiaoqing Zhou

‡

April 2022

Abstract

One of the most important channels through which monetary policy affects the real

economy is changes in mortgage rates. This paper studies the effects of mortgage rate

changes resulting from monetary policy shifts on homeowners’ spending, debt repayment

and defaults. The Canadian institutional setting facilitates the design of identification

strategies for causal inference, since the vast majority of mortgages in the country

experience predetermined, periodic, and automatic contract renewals with the mortgage

rate reset based on the prevailing market rate. This allows us to exploit quasi-random

variation in the timing of the rate reset and present causal evidence for both rate declines

and increases, with the help of detailed, representative consumer credit panel data. We

find asymmetric effects of rate changes on spending, debt repayment and defaults. Our

results can be rationalized by the conventional cash-flow effect in conjunction with

changes in consumer expectations about future interest rates upon the reset. Given the

pervasiveness of Canadian-type mortgages in many other OECD countries, our findings

have broader implications for the transmission of monetary policy to the household sector.

Keywords: Mortgage rate, monetary policy, consumption, consumer expectations,

household finance.

JEL Codes: D12, D14, E43, E52, G21, R31.

Declarations of interest: None.

We thank Jason Allen, James Cloyne, Scott Frame, Kris Gerardi, Lutz Kilian, Jonathan Parker, Tomasz Piskorski, Luigi

Pistaferri and Joseph Vavra for helpful comments and discussions. We also thank the editor and two anonymous reviewers

for their constructive comments. Maria teNyenhuis provided excellent research assistance. The views in this paper are

solely the responsibility of the authors and should not be interpreted as reflecting the views of the Bank of Canada, the

Federal Reserve Bank of Dallas, or the Federal Reserve System.

†

Katya Kartashova, Bank of Canada, 234 Wellington Street, Ottawa, ON K1A 0G9, Canada. Email:

kartashova.katya@gmail.com.

‡

Xiaoqing Zhou, Federal Reserve Bank of Dallas, 2200 N. Pearl St., Dallas, TX 75201, USA. Email: xqzhou3@gmail.com.

1 Introduction

One of the most important channels through which monetary policy aﬀects the real economy is

changes in mortgage rates. Lower mortgages rates, for example, reduce long-term borrowing costs

and debt service expenses, increase borrowers’ cash ﬂows, and stimulate consumer spending in

economic downturns. A persistent challenge in evaluating the mortgage-rate channel has been

the identiﬁcation of exogenous mortgage rate changes and their eﬀects on consumer spending and

savings. Given the diﬃculty of using time series analysis to isolate the mortgage-rate channel from

all other channels through which monetary policy aﬀects aggregate spending and savings, a growing

number of studies have turned to micro data and cross-sectional empirical strategies.

The work by Di Maggio et al. (2017), for example, focuses on U.S. adjustable-rate mortgage

(ARM) borrowers over a declining interest rate period. Their strategy, however, cannot be applied

to studying the causal eﬀects of mortgage rate increases on consumer behavior in the U.S.,

because creditworthy consumers can always reﬁnance their ARMs to avoid adjusting to higher

rates, creating a selection problem in the estimation. Empirical strategies used by other related

studies have relied on comparing ARM borrowers with ﬁxed-rate mortgage (FRM) borrowers (e.g.,

Jappelli and Scognamiglio (2018); La Cava et al. (2016); Floden et al. (2021)), or comparing

mortgage borrowers with outright homeowners (e.g. Agarwal et al. (2022)), which is prone to the

selection-into-treatment concern. In addition, most studies have focused on a single episode around

the global ﬁnancial crisis when major central banks slashed their policy rates.

The reversal of

the low policy rate environment since late 2021 in major economies has created interest in the

corresponding eﬀects of a monetary tightening.

These challenges motivate us to explore a diﬀerent institutional setting, the Canadian mortgage

market, using detailed, representative consumer credit panel data. Canada is an interesting case

to consider, because the institutional setting of the mortgage market permits a clean identiﬁcation

design. The vast majority (about 80%) of mortgages in Canada are “short-term” FRMs that are

subject to predetermined, periodic and automatic contract renewals. Speciﬁcally, the interest rates

on these mortgages are ﬁxed within a term (typically 2-5 years) and have to be reset at the end

of the term based on the prevailing market rate, while the balance is automatically rolled over.

The only other study that provides evidence for a monetary tightening is Agarwal et al. (2022). Unlike the

strategy used by Agarwal et al. (2022) that compares mortgage borrowers with outright homeowners, our comparison

is between similar mortgage borrowers that only diﬀer in the timing of the contract renewal. Moreover, our sample

is nationally and geographically representative of mortgage borrowers, whereas the sample in Agarwal et al. (2022)

is representative of credit card holders.

Borrowers usually renew their contracts with their current lenders, in which case, their credit

scores, loan-to-value (LTV) ratios and debt-to-income (DTI) ratios are not reassessed, unlike when

originating a new mortgage. Moreover, due to large prepayment penalties, most borrowers renew

their contracts and reset their rates as scheduled.

These institutional features imply that the timing of the mortgage rate reset is determined by

past contract choices, and that the change in the mortgage rate upon the reset results from the

change in the prevailing market rate over the contract period, rather than being driven by the

borrower’s ﬁnancial condition, creditworthiness, or spending decisions. Intuitively, we can compare

the responses of two borrowers that are similar in every aspect except that one borrower resets

the mortgage rate earlier than the other borrower. Our identiﬁcation strategy, therefore, exploits

quasi-random variation in the timing of the rate reset in the Canadian mortgage-market setting.

Using this strategy, we provide evidence for two recent monetary policy episodes in Canada that

correspond to major policy shifts during our sample period (Figure 1): the expansionary episode of

2015m1-2017m1 and the contractionary episode of 2017m7-2019m6. We start by showing that these

policy changes were passed through to mortgage rates when borrowers renewed their contracts, with

the magnitude of the pass-through depending on the length of the contract term before the reset.

In the expansionary episode, the rate reduction ranged from 16 basis points (bps) to 113 bps, with

the largest reductions experienced by borrowers who had relatively long terms (e.g., four and ﬁve

years) prior to the reset. In the contractionary episode, renewing borrowers saw rate increases

of 32-85 bps, with the smallest increases experienced by borrowers who had relatively long terms

before the reset. Changes in rates translate to changes in interest payments. We estimate that

the required mortgage payments fell (rose) by $14-$92 ($34-$83) per month in the expansionary

(contractionary) episode, or by $2,907-$20,891 ($7,072-$19,165) over the remaining life of the loan,

if the borrower uses the same length of time to pay it oﬀ.

Given the changes in mortgage rates and payments upon the reset, we examine their eﬀects on

consumer spending using two measures available in our data. One is newly originated auto loans

that capture spending on automobiles. The other is newly originated installment loans (excluding

student loans), which are often used by Canadian consumers to cover large, one-time expenses such

as home improvements and purchases of durable goods (e.g., furniture and consumer electronics).

Standard consumption theory predicts that consumers increase (decrease) their spending when

mortgage payments fall (rise), and that liquidity-constrained borrowers are more responsive to

these cash-ﬂow shocks (e.g., Carroll (1997), Deaton (1991), and Kaplan and Violante (2014)).

Our results for the expansionary episode are in line with this theory: Borrowers receiving the

largest cash ﬂows increased spending on automobiles by 16% and spending ﬁnanced by installment

loans by 18%. Young borrowers and borrowers with higher credit scores are more responsive to

these shocks. In the contractionary episode, however, the types of spending measured in our data

did not change signiﬁcantly, suggesting that consumers are willing to smooth their consumption

to a larger extent in response to negative cash ﬂows. The asymmetric consumption responses to

income shocks have been documented in other contexts. For example, the work by Baugh et al.

(2021) shows that consumers increase their consumption when receiving tax refunds, but these

same consumers do not cut their spending when making tax payments, regardless of whether these

payments are expected or unexpected.

Another important aspect of household responses to mortgage rate changes is the balance-sheet

adjustment. If lower mortgage rates cause households to deleverage on their debt because of

positive cash ﬂows, for example, the resulting strengthened balance sheets will support future

consumption. We therefore examine debt-repayment responses to rate resets, distinguishing

between mortgage prepayment and the paydown of revolving consumer credit, i.e., credit cards

and lines of credit (LOCs). Again, we ﬁnd asymmetric patterns. Borrowers renewing their

contracts in the expansionary episode paid down their mortgage principal faster, by scheduling

higher monthly payments than the required amount for the new term, whereas borrowers in the

contractionary episode did not deviate from their required payments, implying no acceleration in

the amortization. For consumer revolving debt, while we do not ﬁnd a change in the overall balance

in the expansionary episode, we document across-the-board repayments of this type of debt during

the contractionary episode. Taken together, these results imply that consumers deleveraged on

their debt in both episodes.

The asymmetric debt-repayment responses we document are new to the literature. Our ﬁndings

suggest that rate resets aﬀect consumers not only through changes in cash ﬂows, as stressed in

the previous literature, but also through other channels. Otherwise, negative cash ﬂows in the

contractionary episode would drive consumers to accumulate more debt to smooth consumption.

We evaluate a number of potential explanations and provide evidence that changes in consumer

expectations about future interest rates help rationalize our ﬁndings.

The third aspect of household responses we examine is the change in the likelihood of being

delinquent on payments. In line with previous studies (e.g., Fuster and Willen (2017)), we ﬁnd

lower delinquency rates on mortgages and consumer credit products as borrowers reset to lower

rates in the expansionary episode. In the contractionary episode, however, there is no evidence for

rising delinquencies, which may be explained by high expected costs of default and strict lending

standards in Canadian mortgage and consumer credit markets that ensure borrowers’ resilience to

negative cash ﬂows.

Lastly, given our micro-level analysis, we expect rate resets to generate sizable eﬀects on

aggregate spending when borrowers reset their rates to lower levels, and to contribute to aggregate

savings when borrowers experience rate increases. Since our sample is representative of Canadian

borrowers, we are able to provide estimates of the aggregate eﬀects. Our estimates show that,

between 2015m1 and 2017m1, additional auto spending caused by mortgage rate resets amounted

to $2.02 billion, or 1.53% of aggregate new auto sales, while additional durable consumption ﬁnanced

by installment loans amounted to $1.38 billion, or 0.42% of aggregate durable consumption. In the

contractionary episode between 2017m7 and 2019m6, deleveraging on revolving debt upon the rate

reset raised aggregate savings by $693 million, or 1.72% of the aggregate saving.

Relation to the literature. The importance of understanding the transmission of monetary

policy to households has given rise to a large number of studies on the eﬀects of mortgage rate

changes on consumption, balance sheets and defaults (e.g., Di Maggio et al. (2017), Agarwal et al.

(2022), Jappelli and Scognamiglio (2018), La Cava et al. (2016), Floden et al. (2021), Tracy and

Wright (2016), Fuster and Willen (2017), Ganong and Noel (2020), Agarwal et al. (2017), Ehrlich

and Perry (2017), Karamon et al. (2017) and Abel and Fuster (2021)). Our paper falls into this

literature.

Among these studies, our paper is most closely related to Di Maggio et al. (2017) with three

major diﬀerences. First, as discussed earlier, the Canadian setting allows us to provide causal

estimates not only for the eﬀects of declining mortgage rates, as in Di Maggio et al. (2017), but

also for the eﬀects of rate increases during a monetary tightening period. Second, our results are

based on a nationally and geographically representative sample of Canadian mortgage borrowers,

whereas the sample underlying the main results of Di Maggio et al. (2017) is not representative

of the vast majority of mortgages or borrowers in the U.S., as shown in Appendix A. Speciﬁcally,

the jumbo prime interest-only ARMs in their study accounted for only 1.8% of the total U.S.

mortgage origination in 2005-2007, and almost half of this type of mortgages were originated in

California alone. Whether the consumption and payment responses of these ARM borrowers can

be generalized to a typical American consumer who is likely to hold a 30-year FRM is unclear.

Third, as in other related studies, Di Maggio et al. (2017) interpret rate resets as cash-ﬂow shocks,

whereas our paper suggests that rate resets also aﬀect consumer responses through expectations of

future interest rates.

More broadly, our work is related to the literature on consumption and saving responses to

income shocks (e.g., Baugh et al. (2021), Agarwal and Qian (2014), Agarwal et al. (2007), Johnson

et al. (2006), Parker et al. (2013), and Kaplan and Violante (2014)), and the literature on the

transmission of monetary policy shocks to households through housing and mortgage markets and

various other channels (e.g., Kaplan et al. (2018), Cloyne et al. (2020), Beraja et al. (2019), DeFusco

and Mondragon (2020), Wong (2021), Greenwald (2017), Chen et al. (2020), Hurst and Staﬀord

(2004), Bhutta and Keys (2016) and Zhou (2022)). We discuss in more detail the relation between

our work and these studies in subsequent sections.

The remainder of the paper is organized as follows. Section 2 discusses key institutional features

of the Canadian mortgage market, an overview of the Canadian consumer credit market, and

the credit panel data used in our analysis. Section 3 describes the empirical strategy. Section

4 examines the eﬀects of rate resets on mortgage loan-level outcomes. Section 5 examines the

eﬀects on consumer-level outcomes, discusses and evaluates the channels through which resets aﬀect

consumers, and provides evidence from expectations survey data that supports the expectations

channel. Section 6 presents estimates for the aggregate eﬀects of rate resets on spending and debt

repayment. Section 7 provides further evidence and robustness checks. Section 8 concludes.

2 Institutional Setting and Data

2.1 Canadian Mortgage Market

The Canadian mortgage market has several important institutional features that facilitate our

identiﬁcation design. First, unlike the U.S. mortgage market where long-term FRMs are dominant,

the vast majority of Canadian mortgages have short terms (2-5 years) and long amortization periods

(25-30 years).

The amortization period is the length of time it takes to pay oﬀ a mortgage, whereas

the term is the length of time the contract, especially the mortgage rate, is in eﬀect. Having a

Our credit panel data show that 95% of FRMs in Canada have a term of 2, 3, 4 or 5 years. Since FRMs account for

80% of the market, the four products, 2-, 3-, 4- and 5-year FRMs, together account for 76% of the entire Canadian

mortgage market, with 5-year FRMs most popular (accounting for 60% of FRMs). The Bank of Canada-OSFI

Mortgage Origination Dataset conﬁrms the high presence of these four products (73% of total originations since

2014).

mortgage term that is shorter than the amortization period requires the borrower to renew the

contract by the end of the term. Upon the renewal, the remaining balance is rolled over and

the mortgage rate is reset based on the prevailing market rate.

Typically, by the end of the

amortization, a mortgage contract would have been renewed several times.

Consumers are allowed to change the length of the term in the renewal process. In our data,

consumers holding the most common contracts, i.e., 2- and 5-year FRMs, tend to keep their previous

contract length at renewal (see Appendix Table D1). For example, 65% of 2-year FRM borrowers

and 68% of 5-year FRM borrowers who renewed their contracts between 2015 and 2017 chose the

same term length as in the previous term. Other borrowers, such as those holding a 3- or 4-year

contract, either maintained their previous term length or switched to a more common term, 2 or 5

years, that is closer to their previous term length. We defer the discussion of how the switching of

contract terms can potentially aﬀect our empirical results to Section 7.

Second, most borrowers renew their mortgage contracts with their current lender. We estimate

that among borrowers who eventually renewed their mortgages between 2015 and 2017, 98% did so

with their current lender. This share fell slightly to 97% in 2017-2019.

When a mortgage is renewed

with the current lender, the lender in general does not reassess the borrower’s risk measures, such

as the credit score and LTV and DTI ratios. Thus, both rate increases and decreases are passed

on to the borrower automatically. This feature diﬀers from mortgage reﬁnancing in the U.S. that

almost always requires reassessing the underwriting criteria.

Third, the existence of prepayment penalties ensures that borrowers renew their mortgage

contracts as scheduled. Although the penalty varies from lender to lender, it is usually the higher

of (i) three months’ interest on the remaining balance, and (ii) the interest diﬀerential based on the

contract rate and the current market rate for a term of the same length as the remaining time left

on the current term. When the mortgage rate declines, the interest-diﬀerential penalty captures all

ﬁnancial gains from prepaying the mortgage in full and originating a new mortgage at a lower rate.

In practice, borrowers may renew their contract slightly earlier than scheduled without paying a

penalty, provided that their current lender allows them to do so. As shown in Figure 2, more than

Mortgage rates at the borrower level may vary slightly with the borrower’s bargaining power (Allen et al. (2014,

2019)). In our analysis, we control for borrower ﬁxed eﬀects and a set of borrower characteristics, which help remove

the sources of variation in bargaining power.

These shares are computed using information in Appendix Table D2. In the expansionary episode, for example,

the probability of renewing the contract with the current lender conditional on the event that the consumer

eventually renewed her contract is (0.72+0.28*0.1*0.38)/(0.72+0.28*0.1)=0.98. Likewise, this probability is 0.97

in the contractionary episode.

98% of renewals in our data occurred in the six months leading up to the scheduled dates, with

on-time renewals accounting for 50%.

The existence of prepayment penalties also reduces the incentive of Canadian consumers to

extract their home equity through cash-out reﬁnancing, a process that involves paying oﬀ the

current balance and originating a new loan secured by the same property with a higher balance.

Cash-out reﬁnancing is most likely to happen when a consumer approaches her scheduled renewal

date and when prepayment penalties do not apply. Using our data, we identify cash-out reﬁnancing

activity for loans terminated before the scheduled renewal (see Table D2), and then estimate the

annual cash-out reﬁnancing propensity to gauge the quantitative importance of this activity. We

estimate this propensity to be 3% in Canada between 2015 and 2017, which is much smaller than the

7%-11% propensity in the U.S. during the early-2000s reﬁnancing boom, according to Bhutta and

Keys (2016).

Compared to cash-out reﬁnancing, taking out home equity lines of credit (HELOCs)

is not subject to prepayment penalties and is more ﬂexible in the amount that consumers can

borrow, hence more prevalent in Canada (Ho et al. (2019)). We study the eﬀect of rate resets on

HELOC balances in Section 5.

2.2 Canadian Consumer Credit Market

A central part of our analysis is the eﬀects of mortgage rate resets on the balances and payments

of consumer credit that includes revolving debt, such as credit cards and LOCs, and non-revolving

debt, such as auto, installment and student loans.

As in many countries, consumer credit

constitutes an important part of household debt in Canada, provides indicators for changes in

consumer spending and savings, and is closely watched by policymakers assessing ﬁnancial risks.

The Canadian consumer credit market shares some common features with the US market.

First, it accounts for a signiﬁcant share of outstanding household debt in both countries, about

34% in Canada (according to Statistics Canada) and 24% in the U.S. (according to the Federal

Reserve Board) since 2015. The diﬀerence is partially driven by the statistical agencies’ treatment

of HELOC debt, which is classiﬁed as consumer credit in Canada but as mortgage debt in the U.S.

While not fully prepayable, Canadian mortgage contracts allow for an annual prepayment of up to 20% of

the initial balance on top of the scheduled amortization without penalty. This partial prepayment, however, is not

associated with a change in the mortgage rate, and hence does not aﬀect our identiﬁcation.

We estimate the Canadian cash-out reﬁnancing propensity as follows. According to Table D2, among borrowers

scheduled to renew their mortgages, the share of cash-out reﬁnancing borrowers is 13% (=28%*46%). In the

population, the share of borrowers scheduled to renew their contracts among all FRM borrowers is 25% in a year.

Taking these facts together, we estimate the annual cash-out reﬁnancing propensity to be 3% (=25%*13%) over the

2015-2017 period. This propensity is slightly lower in the 2017-2019 period (2.6%), when interest rates moved up.

For the purpose of our study, we exclude student loans, which represent 5% of the overall consumer credit balance

between 2015 and 2019 and are even less important among homeowners.

Excluding HELOC debt, consumer credit accounts for 22% of Canadian household debt, much closer

to the US share. Second, the per capita balances and credit utilization rates on comparable debt

instruments, such as auto loans and credit cards, are similar in the two countries, after accounting

for the exchange rate (by comparing the TransUnion Canada data and the New York Fed Consumer

Credit Panel Reports). These features suggest broad applicability of our estimates to economies

with similar consumer credit markets.

On the other hand, there are some diﬀerences between the two markets. First, apart from

the classiﬁcation of HELOCs, the importance of HELOC debt in overall household debt diﬀers

substantially. Whereas HELOC balances accounted for less than 4% of US household debt after

2015, it captured almost 12% of Canadian household debt over the same period. This is likely

explained by the less prevalent use of cash-out reﬁnancing as a tool for home equity extraction

in Canada as discussed earlier. The second diﬀerence is the lower delinquency rates in Canada

compared to the U.S. The percentage of balances delinquent for more than 90 days in the U.S., across

diﬀerent consumer credit products, was about four to eleven times as high as the corresponding

percentage in Canada over the period we study. This is likely explained by the tight lending

standards in Canadian mortgage and consumer credit markets, as well as the high expected costs

of default in Canada. As we show in Section 5.5, even in the contractionary episode, delinquencies

did not rise for borrowers who resett their mortgage rates to higher levels.

2.3 Data

We use granular account-level (or trade-line-level) data provided by TransUnion Canada, one of the

two credit reporting agencies in Canada, which collects information on 35 million individuals and

covers nearly every consumer in the country that has had a credit report. The data are available

from 2009 onwards at the monthly frequency.

For each consumer, we merge mortgage loan-level

information with consumer-level information on non-mortgage debt that is compiled by ourselves

using account-level data on auto loans, installment loans, credit cards and LOCs. This bottom-up

approach allows us to precisely identify the timing and the amount of a purchase that is ﬁnanced

by an auto loan or an installment loan.

The mortgage loan-level data have information on the origination date, initial amount, insurance

The data collected by TransUnion Canada are reported in accordance with the Metro 2 format of the Canadian

credit reporting guidelines, which specify the variables for reporting. To protect the privacy of Canadians, no

personal information was provided by TransUnion. The TransUnion dataset was “anonymized,” meaning that it does

not include information that identiﬁes individual Canadians, such as names, social insurance numbers or addresses.

In addition, the dataset has a panel structure, which uses ﬁctitious account and consumer numbers assigned by

TransUnion.

status, whether the loan is taken out jointly, and whether the borrower is the primary holder of

the loan, as well as other origination information. The data also include monthly updates on the

current balance, scheduled payment, term duration, delinquency status and whether the loan is

terminated. Information on non-mortgage debt follows a similar data structure. Moreover, the

data provide information about a borrower’s age, credit score, and forward sortation area (FSA,

corresponding to the ﬁrst three digits of a postal code).

For the purpose of our study, knowing the exact timing of a rate reset is crucial. Not all

mortgages in the dataset can be associated with their renewal dates, however, because some lenders

do not report the term duration information, making it impossible to infer the renewal dates of

their mortgages. For this reason, we use mortgages issued by one of the largest commercial banks

and their corresponding consumers as our sample. This bank is the only major bank that reports

the term duration information.

Our sample is representative of the mortgages and borrowers in Canada. First, mortgages

originated by the bank have very similar characteristics to those originated by other federally

regulated lenders (see Table 1). Second, the bank’s share in Canadian mortgage originations and in

the aggregate stock of Canadian mortgages are both close to 20%. Third, the bank operates in all

regions of the country and has a market share of about 20% in each Canadian province, suggesting

the geographic representativeness of our sample.

2.4 Construction of Key Variables

Mortgage rates. Our analysis requires information on the mortgage rate type (i.e., ﬁxed or

variable) and the mortgage rate, which is not provided by the TransUnion data. To identify the

mortgage rate type, we classify a loan as ﬁxed-rate in a term if the scheduled monthly payment

does not change during that term.

We then design a procedure to recover the rates associated

with these FRMs. In Appendix B, we describe in detail how these rates are constructed and how

we use two alternative datasets that contain information on actual mortgage rates to validate our

procedure. We show that the distribution of the imputed rate based on this procedure closely

matches that in the two alternative datasets.

Required monthly payments. Given the new mortgage rate upon the reset, we construct a

payment schedule that is not observed in our data but measures the automatic adjustment of

the monthly payment implied by the rate reset. In constructing this variable, we assume the same

Although some lenders in Canada oﬀer ﬁxed-payment schedules for variable-rate mortgages, the lender in our

sample typically does not.

remaining amortization period and the same outstanding balance as in the month prior to the reset.

Comparing the change in this required payment to the change in the scheduled payment set by the

borrower allows us to examine the choice between mortgage prepayment and cash withdrawal upon

the reset.

Remaining amortization. If a borrower sets a higher monthly payment than the required

amount, the remaining amortization period shortens. For each renewed mortgage, we ﬁrst use

the pre-renewal rate, balance and monthly payment to infer the remaining amortization had the

mortgage not been renewed. We then use the post-renewal rate and monthly payment and the

pre-renewal balance to infer the amortization after the reset.

Spending measures. We construct two measures for durable spending. First, we use newly

originated auto loans as proxies for spending on automobiles.

Since we work directly with

loan-level data, we can precisely identify the timing and the amount of a vehicle purchase that

is ﬁnanced by an auto loan. Second, we use newly originated installment loans (excluding student

loans) to measure broader types of durable spending. In Canada, these loans are designed to cover

large one-time expenses and are typically used for home improvements and purchases of furniture

or other durable goods.

Classifying HELOCs. The original account-level data do not distinguish between HELOCs,

which are secured by borrowers’ homes, and unsecured LOCs. While both types of LOCs are

revolving debt products, the former have lower rates and provide a more important source for

homeowners to smooth their consumption.

Hence, we are particularly interested in understanding

changes in HELOC balances upon the rate reset. We follow the industry practice in classifying

HELOCs as LOCs that have initial credit limits of at least $50,000. Our empirical results are robust

to changing this threshold.

Delinquency measures. Given loan-level information on the delinquency status, we create

consumer-level measures of delinquency on each type of debt as follows. First, we create an indicator

at the loan level that takes the value of one if the loan in the current month approaches a certain

level of delinquency (60 or 90 days). We then count the number of newly delinquent accounts for

each type of debt. Finally, we convert the number of newly delinquent accounts into a dummy

According to Watts (2016), as of 2016, 83% of new motor vehicles in Canada were obtained with ﬁnancing, and

the trend of ﬁnanced vehicle sales has closely tracked that of total sales. In addition, historical data show that the

average LTV ratio of motor vehicles in Canada is close to 100%.

Bailliu et al. (2012) estimate that 40%-50% of the funds extracted from home equity is used for consumption

and home improvement in Canada.

variable that indicates new delinquency on at least one account of a given debt category.

Constructing LTV ratios. Since the credit panel data do not provide borrower-level information

on house prices, we instead construct the FSA-level LTV ratio and include its lag as a control

variable in our regressions. Speciﬁcally, we use the FSA-level median house price in 2014Q1 (from

the Bank of Canada-OSFI mortgage originations dataset) multiplied by the growth rate of the

Teranet quarterly FSA-level house price index to obtain the FSA’s current-quarter house price. We

then divide the median mortgage balance of the FSA’s consumers by the imputed current-quarter

house price of the FSA to obtain the LTV ratio.

2.5 Summary Statistics

Table 2 shows the summary statistics of key variables at the mortgage loan level and at the consumer

level.

We perform the analysis on loans renewed in 2015m1-2017m1 for the expansionary episode,

and on loans renewed in 2017m7-2019m6 for the contractionary episode. For tractability, we restrict

our analysis to FRMs that have (pre-renewal) terms of 2, 3, 4 or 5 years, which jointly account for

95% of the FRM market. Summary statistics are presented for each term separately. Our analysis

is restricted to primary mortgage holders. Borrowers that have more than one mortgages at the

same time are excluded. In total, we have 88,328 loans renewed during the expansionary episode

and 85,376 loans renewed during the contractionary episode.

3 Empirical Strategy

Our empirical strategy is designed to exploit variation in the predetermined timing of mortgage

rate resets in the two episodes. In essence, we compare the responses of two borrowers who are

similar in every aspect except that one borrower resets her mortgage rate earlier than the other

borrower. Our analysis is carried out separately for each mortgage term in each episode, so we can

focus on borrowers with similar contracts and avoid the potential concern of selection into speciﬁc

terms.

The panel structure of our data allows us to include a set of borrower-level characteristics,

borrower ﬁxed eﬀects, and month-of the-sample ﬁxed eﬀects that may confound the eﬀects of rate

resets. Our baseline speciﬁcation is

j,t

= γ

+ δ

+ α

P ostRenew

j,t

+ α

j,t

+ ε

j,t

, (1)

where y

j,t

is an outcome of borrower (or loan) j in month t. P ostRenew

j,t

is the indicator for

To limit the eﬀect of outliers in our data, we eliminate the top 1% observations for balance and payment variables.

Our empirical results are robust to using alternative thresholds such as the top 0.5%, 2.5% and 5%.

the months after borrower j’s contract renewal. x

j,t

is a vector of the borrower’s characteristics,

including the lagged credit score, age and the previous-quarter FSA-level LTV ratio. γ

the borrower ﬁxed eﬀect that absorbs unobserved heterogeneity potentially correlated with the

borrower’s choices. δ

is the month-of-the-sample ﬁxed eﬀect designed to capture the trend in the

aggregate economy and to control for the confounding eﬀects of aggregate shocks. α

is the key

parameter of interest that captures the eﬀect of the rate reset. The standard errors are clustered

at the borrower level.

Economy theory suggests that consumers may respond diﬀerently to the same mortgage rate

shocks due to heterogeneity in their wealth and access to credit markets. We consider three

empirical measures to quantify these heterogeneous responses. First, we use the average credit

score over the preceding 12 months as a proxy for the borrower’s access to new credit. Second,

we use the average utilization rate of revolving credit over the preceding 12 months as a proxy for

constraints on existing, available credit. Third, we use ages under 45 or above 65 as a proxy for low

liquidity levels compared to middle-aged consumers. This choice is consistent with the prediction of

standard life-cycle theory and the patterns in household survey data (e.g., the Survey of Consumer

Finances and Panel Study of Income Dynamics). We interact each of these empirical measures with

the post-renewal indicator, P ostRenew

j,t

, for the estimation of heterogeneous eﬀects.

Consumers may change their spending and savings before the reset in anticipation of a rate

change upon the reset. To evaluate the importance of this anticipation eﬀect, we estimate the

dynamic version of equation (1) that includes a set of quarterly dummies to replace the post-renewal

indicator. Speciﬁcally, we estimate the α

’s from

j,t

= γ

+ δ

q∈Q

(t ∈ q) + α

j,t

+ ε

j,t

, (2)

where 1

(t ∈ q) is an indicator equal to one if month t is in the qth quarter since the mortgage

renewal. We set the quarter before the renewal as quarter zero and estimate the responses in the

three quarters before and ﬁve quarters after the renewal relative to quarter zero.

One potential concern with our baseline strategy is its inability to account for the mortgage-age

eﬀect. For example, consumers may be less likely to buy a car at the time when they buy a house,

perhaps because they have used up their savings to make the mortgage down payment, or because

Our results are robust to including the province-by-quarter ﬁxed eﬀects or the cohort-by-quarter ﬁxed eﬀects.

The ﬁrst set of ﬁxed eﬀects controls for region-speciﬁc time trends. For example, the eﬀect of oil price shocks may

vary substantially across regions, as the oil sector is geographically concentrated in Canada (see Kilian and Zhou

(2020)). The second set of ﬁxed eﬀects controls for unobserved heterogeneity across cohorts.

they are too leveraged to be qualiﬁed for a new auto loan. As consumers accumulate savings after

the home purchase, they are more likely to buy a car. This means that consumers’ auto spending

increases with the age of their mortgage. The existence of this mortgage-age eﬀect threatens our

identiﬁcation, because we might attribute the increase in auto spending to the eﬀect of the rate

reset. Ideally, we would include a set of dummy variables indicating the age of the mortgage. These

variables would be collinear with the post-renewal indicator, so we cannot control for them.

We therefore implement two alternative diﬀerence-in-diﬀerence (DID) strategies for robustness.

First, we consider a design that introduces FRMs with long terms (7 or 10 years) as the control

group. These mortgages were previously renewed at the same time as mortgages in our sample.

Essentially, this strategy compares two mortgages that were both previously renewed in, say,

2010m1. One was renewed again in 2015m1, while the other had to wait for another two years.

In the second design, we introduce as the control group mortgages having the same terms as the

treatment group but not renewed in the episode. For example, in the expansionary episode, we

use 5-year FRMs previously renewed in 2012m1-2013m1 (hence not renewed in the expansionary

episode) as the control group for 5-year FRMs renewed in this episode. This approach mitigates

the selection concern arising from comparing diﬀerent types of mortgage contracts.

Both designs can be implemented by estimating the following speciﬁcation:

j,t

= γ

+ δ

+ β

Renew

× P ostRenew

j,t

+ β

j,t

+ ε

j,t

, (3)

where Renew

is the indicator for loan j to be renewed in an episode. Other variables are similarly

deﬁned as in equation (1). The parameter of interest is β

, which captures the DID eﬀect. As we

discuss in Section 7, the estimates using the baseline strategy are similar to these DID estimates.

4 Mortgage Loan-Level Adjustments

We start by estimating the changes in the mortgage rate and required monthly payment upon the

reset, and then turn to the change in the monthly payment scheduled by the borrower. The latter

reﬂects the borrower’s choice of whether to pay down the mortgage principal faster by deviating

from the required payment. Heterogeneity in this choice across borrowers shows support for the

prediction of standard consumption theory.

4.1 Changes in Mortgage Rate and Required Payment

Column (1) of Table 3 shows the change in the mortgage rate for each term. In the expansionary

episode, borrowers renewing their mortgages experienced substantial downward adjustment in the

rate. The magnitude, however, depends on the term prior to the reset. Mortgages with relatively

longer terms had larger declines in rates. For example, the average decline was 113 bps for 5-year

FRMs, but 16 bps for 2-year FRMs. This diﬀerence is due to the fact that the prevailing market

rate had already been declining before the episode started, leading to larger cumulative changes for

mortgages with longer terms.

Lower mortgage rates imply savings on required interest payments. In column (2), we estimate

the change in the required monthly payment, which quantiﬁes the maximum payment reduction

per month the borrower can realize in the new term. Consistent with the rate-change pattern,

5-year FRM borrowers had the largest reduction in the new required payment, $92 per month on

average, whereas the reduction for 2-year FRM borrowers was only $15 per month. Given the

remaining amortization for each type of mortgage, we estimate the total interest savings upon the

reset, assuming that the amortization and mortgage rate do not change. This ranged from a modest

amount of $2,907 to a substantial amount of $20,891, depending on the pre-renewal term (see Table

4). As we show in Section 4.2, borrowers also shortened their amortization upon the reset, implying

further interest savings.

In the contractionary episode, borrowers renewing their mortgages experienced rate increases.

The magnitude decreases with the term prior to the reset, due to the reversal of the declining rate

trend. Borrowers with 2-year FRMs, for example, experienced the largest rate increase of 85 bps,

whereas borrowers having 5-year FRMs encountered a 32-bp rate increase on average. We estimate

that the monthly required payment increased by $34-$83, depending on the pre-renewal term.

Assuming the amortization and current rate unchanged, the total increase in interest payments

upon the reset in this episode ranged from $7,072 for 5-year FRM borrowers to $19,165 for 2-year

FRM borrowers.

These estimates represent the average changes experienced by borrowers who renewed their

mortgages in an episode. At the individual level, however, the changes may diﬀer depending

on the timing of the renewal. In the contractionary episode, for example, the prevailing market

rate increased gradually following the policy rate movements (Figure 1), implying that borrowers

renewing their mortgages early in the episode experienced smaller increases in rates and payments

than borrowers renewing their mortgages later, even though they held the same type of contract.

Our estimates put more weight on mortgages renewed early in an episode (due to more observations

of these loans in the post-renewal period), which could cause us to underestimate the true average

increases in rates and payments.

To address this concern, we provide estimates using the sample that restricts the post-renewal

observations to be within four quarters for each loan. The results are shown in Appendix Table D4.

For the contractionary episode, the average increases in rates and payments are indeed larger than

the baseline estimates, with diﬀerences of 20 bps in rates and $20 in required payments. For the

expansionary episode, the average decreases in rates and payments are smaller than the baseline

estimates. These loan-level diﬀerences, however, do not aﬀect much our borrower-level outcomes,

i.e., durable consumption and debt repayment, as shown in Section 5.

4.2 Changes in Scheduled Payment and Amortization

When renewing the mortgage contract, a borrower, given the new rate, may choose a diﬀerent

monthly payment from the required amount. In principle, the payment chosen by the borrower must

not fall below the required level. Scheduling a higher monthly payment than the required amount

allows the borrower to pay down the mortgage principal faster and to shorten the amortization.

Comparing the change in the required payment to the change in the scheduled payment, therefore,

allows us to infer the borrower’s decisions on mortgage prepayment and liquidity withdrawal.

In the expansionary episode, we ﬁnd that, indeed, borrowers did not set their new payments

to the low levels required by lenders (Table 3, column 3). For example, 5-year FRM borrowers on

average only lowered their monthly payments by $46, despite the maximum possible reduction of

$92 per month. A similar pattern is found for other types of renewing mortgages. This implies that

only part of the interest savings were realized, and that the rest were used to pay down the principal

faster. How much faster? We estimate the change in the remaining amortization in column (4) of

Table 3. Depending on the term, the amortization period was shortened by 4-14 months. Taking

this into account, we estimate that total interest savings upon the reset ranged from $4,830 to

$23,925 for renewing borrowers (Table 4).

In the contractionary episode, we ﬁnd that borrowers set their new monthly payments to the

levels required by the lenders, leaving the amortization essentially unchanged. The asymmetric

responses of scheduled payments are not surprising, given that lenders in general do not allow

borrowers to schedule a payment lower than required, nor do they permit extensions of the

amortization.

Another way of paying down the mortgage principal is to make a large, one-time payment at

the point of the renewal. The data, however, do not seem to support the prevalence of such large

prepayments upon the reset. We ﬁnd that, only 3.7% of borrowers reduced their balances by more

than 5% upon the reset in the expansionary episode, and that this fraction dropped to 3.2% in the

contractionary episode. In addition, the fraction of borrowers who reduced their balances by more

than 20% is at most 1% in each episode.

4.3 Heterogeneity in Mortgage Payment Choices

We showed that, in the expansionary episode, borrowers used part of their interest savings from the

reset to pay down their mortgage principals faster. There are reasons to believe that this pattern

varies across borrowers. Standard consumption theory predicts that liquidity-constrained borrowers

would allocate more of their interest savings to spending and less to debt prepayment. Here we

focus on borrowers who experienced large payment declines (i.e., 4- and 5-year FRM borrowers) and

examine whether constrained borrowers responded diﬀerently from others. Liquidity constraints

are measured using the credit score, credit utilization and age, as described in Section 3.

Table 5 supports the theoretical prediction. We compute the ratio of the change in

the scheduled payment to the change in the required payment as a measure of liquidity

realization. Consider two groups of borrowers who renewed their 5-year FRMs, one with high

credit scores and the other with low credit scores. Our estimates show that the liquidity

realization ratio for high-credit-score borrowers is 37% (=30.02/82.24), whereas it is 65%

(=[30.02+32.90]/[82.24+14.09]) for low-credit-score borrowers. Similarly, this ratio is 37% for

borrowers with low credit utilization and 68% for borrowers with high credit utilization. Likewise,

this ratio is 57% for young borrowers, 47% for middle-aged borrowers, and 58% for old borrowers.

Turning to borrowers renewing 4-year FRMs, we ﬁnd the similar pattern that liquidity-constrained

borrowers converted more of their interest savings to liquidity than other borrowers.

In contrast, we do not ﬁnd heterogeneity in this ratio in the contractionary episode (not shown

to conserve space). Borrowers set their scheduled payments to the required levels and left the

amortization unchanged, regardless of the liquidity measure. The lack of heterogeneity in the

payment choice in this episode may again be explained by lenders’ policy that the scheduled payment

in general cannot fall below the required level.

5 Mortgage Rate Resets and Consumer-Level Responses

This section examines the eﬀects of rate resets on consumer spending, debt repayment, and defaults.

We ﬁnd interesting asymmetric responses that cannot be fully explained by the cash-ﬂow channel

of the rate reset. This motivates the discussion of alternative channels. We evaluate a number of

possible explanations and provide evidence for the consumer expectations channel using household

survey data.

5.1 Consumer Spending

We ﬁnd that consumers experiencing the largest rate and payment reductions, i.e., 5-year FRM

borrowers, increased their durable spending signiﬁcantly in the expansionary episode. On average,

monthly auto spending and spending ﬁnanced by installment loans rose by $19 and $44, respectively,

equivalent to a 16% and an 18% increase relative to the sample mean (Table 6, columns 1 and 3).

The results also show that rate resets led some consumers who otherwise would not have spent on

these goods to do so. For example, the likelihood of purchasing an automobile increased by 7 bps

in a month (column 2), equivalent to a 19% increase from the mean; the likelihood of taking a new

installment loan increases by 14 bps in a month (column 4), equivalent to a 15% increase from the

mean.

We also address the question of whether these borrowers may have already raised their spending

before the reset, and whether their spending was completely reversed after the initial increase. The

dynamic responses in Figure 3 show that sharp increases in auto and installment-loan-ﬁnanced

spending occurred in the quarter of the reset. The two types of spending remained high for the

next few quarters. In ﬁve quarters, total spending reached $400 for automobiles and $500 for that

ﬁnanced by installment loans, suggesting somewhat lasting eﬀects of rate resets on spending.

Before turning to the contractionary episode, we examine spending heterogeneity, focusing on

5-year FRM borrowers. Standard consumption theory predicts that liquidity-constrained borrowers

would be more responsive to positive cash ﬂows. On the other hand, since interest savings are not

realized immediately but over the course of the new term, diﬃculties in obtaining new credit may

create a hurdle for some consumers who would have borrowed to ﬁnance their current spending.

Our ﬁndings can be summarized into three points (see Table 7). First, while all borrowers increased

their spending upon the reset, low-credit-score borrowers were less responsive, suggesting that they

may have limited access to new credit or face high borrowing costs. Second, young borrowers are

more responsive than other age groups, consistent with the theory. Third, there is no signiﬁcant

heterogeneity across credit utilization, which may be explained by the inability of high-credit-usage

borrowers to obtain more credit.

We do not ﬁnd signiﬁcant spending responses of borrowers renewing other terms of mortgages. This is not

surprising, given that the size of the rate reduction is small for these borrowers.

In the contractionary episode, interestingly, we do not ﬁnd decreases in spending. In fact, with

one exception, spending did not change signiﬁcantly. The only exception is the change in the

likelihood of auto purchases for 2-year FRM borrowers, which is positive, not negative. Nor did

spending decrease at longer horizons (5 quarters). Among borrowers with the same type of contract,

the only noticeable heterogeneity is that low-credit-score borrowers reduced their spending relative

to others, conﬁrming the role of credit market access in explaining spending divergence.

The lack of spending responses in the contractionary episode raises the question of whether

our measures of spending are inadequate for capturing overall consumption responses to negative

cash ﬂows. This is possible, given that our data only measure durable spending ﬁnanced by auto

or installment loans. However, asymmetric consumption responses have also been documented in

other studies using more comprehensive consumption measures. Baugh et al. (2021), for example,

using detailed account-level data on spending made by credit cards, debit cards, and checking and

savings accounts, ﬁnd that consumers respond asymmetrically to positive and negative cash ﬂow

shocks. Speciﬁcally, while consumers increase their consumption upon receiving tax refunds, these

same consumers do not cut their spending when making tax payments in other years, regardless of

whether these payments are expected or unexpected.

5.2 Revolving Debt Balances

Previous studies have documented debt repayment as an important use of positive cash ﬂows (e.g.,

Di Maggio et al. (2017), Bhutta and Keys (2016), and Baugh et al. (2021)). Our results in Section

4.2 showed that consumers pay down their mortgage principal faster when resetting to lower rates,

in line with these studies. We now examine the responses of the balances on revolving credit, i.e.,

credit cards and LOCs.

In the expansionary episode, we ﬁnd that consumers on average paid down their credit card

debt by about $130-$250 upon the reset (or 3%-6% of the mean balance), as shown in column (6)

of Table 6. Deleveraging on credit card debt, however, was completely reversed by rising LOC

balances (column 7), and in particular by higher HELOC balances (column 8). The rise in HELOC

balances indicates that homeowners extract their home equity in response to lower rates (see Bhutta

and Keys (2016)). As a result, the total revolving balance did not change signiﬁcantly (column

5). Figure 4 shows the dynamic responses of credit card and LOC balances for 2- and 5-year FRM

One may also be concerned that the insigniﬁcant estimates are driven by the lack of power of our tests. Recall that

we ﬁnd strongly statistically signiﬁcant spending responses in the expansionary episode for 5-year FRM borrowers.

For the contractionary episode, we apply exactly the same estimation strategy on the sample of a similar size, so the

insigniﬁcant spending responses in this episode cannot be attributed to a lack of power.

borrowers. Credit card balances fell sharply in the quarter of the reset and stayed roughly ﬂat for

the next few quarters, whereas LOC balances rose gradually.

The change in the revolving balance displayed substantial heterogeneity across borrowers. Panel

I of Table 8 shows the results for 5-year FRM borrowers (with similar patterns observed for other

renewers). Overall, high-credit-score, low-credit-usage and old borrowers deleveraged more. The

fact that low-credit-score and high-credit-usage borrowers deleveraged less, sometimes even raising

their leverage, suggests that these borrowers may rely more on existing credit to smooth their

consumption in response to positive cash ﬂows. This can also be seen from the relatively smaller

reduction in their credit utilization rate compared to other borrowers.

Turning to the contractionary episode, our ﬁndings for the responses of revolving debt balances

to rising mortgage rates and payments are novel to the literature. We ﬁnd that consumers paid

down their credit card debt by about $210-$270 upon the reset, equivalent to 5%-6% of the average

balance (Table 6, column 6). Unlike in the expansionary episode, credit card deleveraging was not

oﬀset by rising LOC balances (column 7). Nor do we observe equity extraction through HELOCs

(column 8). Even over longer horizons, LOC balances did not rise (Figure 4, panel b). As a result,

the revolving balance fell by $260-$900 upon the reset, except for 4-year FRM borrowers who

displayed no signiﬁcant change. Among borrowers with the same type of contract, high-credit-score,

low-credit-usage and old borrowers deleveraged more, similar to the patterns observed for the

expansionary episode (Table 8).

5.3 Other Channels of Rate Resets

While our results at the consumer level for the expansionary episode support the conventional

interpretation of rate resets being cash-ﬂow shocks, the ﬁndings for the contractionary episode

reveal that other channels are at work in driving consumer responses. This is because, if the

cash-ﬂow eﬀect dominates when rates increase, consumers are expected to cut their spending or

to raise their debt to smooth the eﬀect of negative cash ﬂows, according to standard consumption

theory. In contrast, we ﬁnd spending unchanged and revolving balances reduced. We consider

several potential explanations of this fact and evaluate their plausibility.

Regarding heterogeneity in the age dimension, ﬁrst, we ﬁnd that old borrowers deleveraged more than other

borrowers on both credit cards and LOCs, consistent with the prediction of life-cycle theory. Second, young borrowers

deleveraged less on credit card debt than other borrowers, but more on LOCs than middle-aged borrowers. The second

ﬁnding may be explained by the fact that young borrowers have high LOC utilization (90%), compared to credit-card

utilization (55%).

The fact that rate increases led to higher mortgage payments, that consumer spending was unchanged, and

that revolving debt was paid down means that consumers must have reduced their savings to ﬁnance these increased

payments. We do not observe household assets in our data, but evidence in Baugh et al. (2021) supports this point

First, one may argue that changes in revolving debt balances simply reﬂect changes in

expenditures not captured by our data (e.g. nondurables and services), and that the reduction

in these balances means lower consumption. This would support the view that negative cash ﬂows

are the main channel through which rate resets aﬀected consumers in the contractionary episode.

This explanation, however, is challenged by two facts. First, it cannot explain why consumers who

experienced the largest negative cash-ﬂow shocks (i.e., 2-year FRM borrowers) did not deleverage

more than borrowers who experienced the smallest cash-ﬂow losses (e.g., 5-year FRM borrowers).

Second, it implies that liquidity-constrained borrowers would cut their spending more and hence

reduce their debt balances more than other borrowers, which is the opposite to what we ﬁnd in the

data.

Second, lenders may force borrowers to deleverage on revolving debt when borrowers reset their

mortgage rates to higher levels, due to the concern that borrowers’ repayment ability may be

undermined by rising mortgage payments. We test this hypothesis by estimating changes in banks’

credit supply to consumers in response to rate resets in the contractionary episode. Columns (1)

and (2) of Table 9 show the change in the likelihood of experiencing a more than $1,000 increase in

the credit limit. This likelihood did not fall, but increased, for credit card debt and for LOC debt.

Columns (3) and (4) show the dollar changes in these credit limits, conﬁrming the patterns in the

previous two columns. Lastly, we estimate the change in the required-payment-to-balance ratio,

which reﬂects the interest rate on the debt (especially for interest-only products). We do not observe

any signiﬁcant change in this ratio (columns 5 and 6). To summarize, we do not ﬁnd supporting

evidence for deleveraging driven by the lender-side tightening of credit in the contractionary episode.

Third, we consider a monthly debt-service-ratio hypothesis that predicts that consumers act to

maintain a constant debt service ratio, i.e., the ratio of monthly debt service payments over monthly

income. Testing this hypothesis is equivalent to testing the hypothesis of constant debt service

payments, provided that consumers’ monthly income does not change.

When mortgage payments

rise, for example, the latter hypothesis suggests that consumers pay down their non-mortgage debt

to keep their total debt service expenses unchanged. We test this hypothesis by estimating the

change in monthly total debt service payments (on mortgages, auto and installment loans, and

revolving debt) to the rate reset. The null hypothesis is no change. The results are shown in

by showing that consumers make account transfers, rather than reduce their consumption, in response to net income

losses resulting from tax payments.

The two hypotheses are equivalent in our setting because our empirical strategy exploits variation in the timing

of the rate reset which is exogenous to changes in consumers’ income (e.g., labor, government transfers, etc.).

Appendix Table D3. We ﬁnd that total monthly payments moved closely with monthly mortgage

payments, with the diﬀerence mainly driven by payments on revolving debt. In the contractionary

episode, in particular, the declines in revolving debt expenses were too small to oﬀset the rising

mortgage payments. Hence, we do not ﬁnd strong support for this hypothesis.

Lastly, we consider an explanation which postulates that, when consumers reset their mortgage

rates to higher levels in the contractionary episode, they expect interest rates to increase in the

future as well.

Since revolving debt often has variable rates, the expectations about rising

future rates drive consumers to pay down this type of debt. Our credit panel data do not

contain information on expectations, so we employ an alternative dataset for the expectations

of representative Canadian consumers to evaluate this hypothesis in Section 5.4.

This expectations channel is closely related to the standard intertemporal-substitution (IS)

channel that predicts that consumers pay down their debt (or equivalently, increase their net

savings) when the current interest rate rises. The expectations channel generalizes the standard IS

channel in two dimensions. First, it links the deleveraging response not only to the current interest

rate but also to the expected future interest rates. Second, it does not necessarily imply symmetric

interest-rate expectations or symmetric deleveraging responses, consistent with our ﬁndings.

5.4 Consumer Expectations

We use the Canadian Survey of Consumer Expectations (CSCE) to evaluate the expectations

channel in explaining consumer deleveraging in the contractionary episode. The CSCE data are

collected every quarter since 2014Q4 from a nationally representative household sample. The

survey asks questions related to households’ ﬁnancial conditions and their expectations about

macroeconomic variables. A detailed description of the data, the survey questions underlying

our analysis, and key summary statistics can be found in Appendix C.

For comparability with our consumer credit panel data, we restrict the CSCE survey sample

to homeowners with mortgages. We establish two sets of results using the CSCE data that jointly

The fact that borrowers holding 2-year FRMs were more likely to switch to longer terms in the contractionary

episode than in the expansionary episode suggests that these borrowers may expect rates to rise for an extended

period (Appendix Table D1).

These diﬀerences are driven by the standard assumption in rational expectations dynamic equilibrium models

that interest rates are constant over time. This assumption gives the standard Euler equation: u

) = β(1 +

r)E

t+1

) = ... = β

(1 + r)

t+T

). By construction, the expected future rates are equal to the current rate

and are symmetric in these models. In contrast, a more general model that relaxes the constant-rate assumption

implies the Euler equation: u

) = β(1 +r

t+1

) = ... = β

(1+r

T −1

j=1

(1 + r

t+j

t+T

)

, which shows

that both the current rate and the expectations about future rates matter for the current saving decisions and that

expectations may not be symmetric.

support the expectations channel. First, consumers who perceive that interest rates have risen over

the past 12 months tend to expect the interest rates to be even higher for an extended period (at

least 5 years). Second, in response to their expectations about rising future rates, consumers are

more likely to pay down debt, cut spending and save more. These patterns persist in the overall

sample and, in particular, in the contrationary episode.

To establish the ﬁrst fact, we estimate the following ﬁxed-eﬀect linear probability model (LPM)

that relates consumers’ perception of recent interest rate changes to their expectations of the rate

in the next 12 months,

ExpectH

i,t

= γ

+ δ

+ θ

CurrentH

i,t

+ x

i,t

+ 

i,t

, (4)

where ExpectH

i,t

is an indicator equal to 1 if consumer i in quarter t expects the average interest

rate to be higher in the next 12 months. CurrentH

i,t

is an indicator equal to 1 if the consumer

perceives that interest rates have risen over the past 12 months. x

i,t

is a vector of consumer

characteristics (i.e., age, gender, marital status and education). γ

and δ

are consumer and time

ﬁxed eﬀects. The standard errors are clustered at the consumer level. Column (1) of Table 10

shows that consumers who perceive that interest rates have risen are more likely to expect rates to

be higher in the next 12 months.

We next consider expectations at longer horizons. The survey asks consumers about their

expected levels of interest rates in one, two and ﬁve years, which allows us to construct indicators

for the rising path of expected future rates. Speciﬁcally, we construct an indicator for expecting

rates to rise in the next two years, which equals one if the consumer (i) expects rates to be higher

in the next 12 months, and (ii) reports the level of the expected rate in two years greater than

the level of the expected rate in one year. Likewise, an indicator for expecting rates to rise in the

next ﬁve years is constructed, which equals one if the consumer meets conditions (i) and (ii), and

reports the level of the expected rate in ﬁve years above the level of the expected rate in two years.

The results in columns (2) and (3) of Table 10 show that consumers who perceive current rates to

have risen are also more likely to expect future rates to increase for an extended period.

To establish the second fact, we take advantage of the survey question that asks consumers

about the actions they are taking or plan to take in response to their interest rate expectations,

including paying down debt, cutting spending/saving more, postponing major purchases, and

bringing forward major purchases. Respondents can choose multiple actions, so we estimate a

series of LPMs similar to equation (4) with one action being a dependent variable at a time.

Columns (4) to (7) in Table 10 show that consumers who perceive rates to have risen are more

likely to pay down their debt, as well as to cut their spending/save more, providing direct support

to the expectations channel in explaining deleveraging in the contractionary episode.

For the expectations channel to explain our earlier results based on the consumer credit panel

data, we have to argue that rate resets must have helped consumers to better understand the

current interest rate, which then triggers changes in their expectations about future rates. This is

not unreasonable, given that mortgage renewals are one of the most important ﬁnancial decisions

for consumers that require their attention. Moreover, interactions with the lender in the process

are likely to draw consumers’ attention to the current market rate. At other times, consumers are

less likely to pay attention to movements in interest rates, as inattention in general is a well known

problem in the household ﬁnance literature (e.g., Keys et al. (2016), Andersen et al. (2020) and

Agarwal et al. (2016)).

We use the CSCE data to provide some evidence for the eﬀect of rate resets on household

expectations. Since the survey does not ask questions about mortgage renewal, we use the indicator

for having a variable-rate mortgage, V RM

i,t

, as a proxy for frequent rate resets, and the indicator

for taking out a new mortgage within the past 12 months, New

i,t

, as a proxy for recent rate resets,

to predict the likelihood of (correctly) perceiving the recent trends in interest rates. We estimate

the following ﬁxed-eﬀect LPM regression, with the omitted group being FRM borrowers who have

taken out the mortgage for more than one year,

CurrentH

i,t

= γ

+ δ

+ η

V RM

i,t

+ η

New

i,t

+ x

i,t

+ 

i,t

. (5)

The results support our hypothesis. For the contractionary episode, the point estimates, ˆη

0.1 and ˆη

= 0.07 (both strongly statistically signiﬁcant at the 1% level), imply that borrowers that

frequently or recently reset their mortgage rates are more likely to perceive interest rates to have

risen. For the expansionary episode, the point estimates, ˆη

= −0.03 and ˆη

= −0.01, suggest that

these borrowers are also more likely to perceive interest rates to have fallen over this episode.

One more piece of evidence supporting the expectations channel is that the average consumer

does not seem to have symmetric expectations about future interest rates. Appendix Table C1

shows that, in the contractionary episode, the majority of consumers perceived rates to have risen

(76%), and that 81% of these consumers expected rates to be higher in the next year. In contrast,

in the expansionary episode, while the majority of consumers perceived rates to have fallen or

unchanged (82%), only 49% of them expected rates to be lower or unchanged in the next year.

These patterns suggest that, while expectations played a key role in explaining deleveraging in

the contractionary episode, they may not have a symmetric eﬀect on debt accumulation in the

expansionary episode, which is consistent with our ﬁndings in the credit panel data.

5.5 Mortgage and Consumer Credit Delinquency

Previous studies using U.S. data have found that lower mortgage rates help reduce mortgage

defaults, shedding light on the policy design that aims to support households in the aftermath

of the global ﬁnancial crisis.

As major economies start to move away from the low policy rate

environment, additional research on the impact of rising interest rates on credit defaults is called

for.

There has not been a causal-inference based study of this issue, however. One challenge in using

U.S. data is that, before the ﬁnancial crisis, ARM resets in the U.S. almost always led borrowers

to increase their monthly payments, and many borrowers responded by reﬁnancing (Fuster and

Willen (2017)). This introduces a selection problem, because borrowers in poor credit conditions

are unlikely to reﬁnance. Thus, comparing ARM borrowers who end up resetting to higher rates

(i.e., who are unable to reﬁnance) with ARM borrowers who are still in the initial rate-ﬁxation

periods could overestimate the eﬀect of rate increases on default. In fact, this selection problem

poses an identiﬁcation challenge not only for estimating the eﬀect on default, but also for the

spending and savings responses using ARM resets.

Focusing on the Canadian mortgage borrowers allows us to circumvent this problem, as discussed

in Section 2.1. The results are shown in Table 11. In the expansionary episode, we ﬁnd that

lower mortgage rates through the reset reduced mortgage delinquencies, especially for 5-year FRM

borrowers, in line with studies using U.S. data. We also ﬁnd that delinquencies on other types of

debt fell for these borrowers. However, we do not ﬁnd that delinquencies change signiﬁcantly for

other renewing borrowers.

In the contractionary episode, we do not ﬁnd that delinquencies increase for borrowers resetting

their rates to higher levels. This is consistent with our earlier ﬁnding that consumers deleveraged on

revolving debt over this period. It may seem surprising that higher mortgage payments did not cause

rising delinquencies. One explanation is that strict lending standards in Canada ensure borrowers’

resilience to negative cash-ﬂow shocks ex post. For mortgages, for example, the regulatory and

supervisory framework has successfully kept mortgage delinquencies at very low levels even during

See, e.g., Tracy and Wright (2016), Fuster and Willen (2017), Ehrlich and Perry (2017), Agarwal et al. (2017),

Ganong and Noel (2020), Karamon et al. (2017), and Abel and Fuster (2021).

the global ﬁnancial crisis, with the highest arrears rate being only 0.45% in January 2011 according

to the Canadian Bankers Association. After the global ﬁnancial crisis, the government tightened

mortgage qualiﬁcation rules, further pushing down the delinquency rate. Another explanation is

the high expected cost of default. For example, almost all mortgages in Canada have recourse

provisions (Crawford et al. (2013)).

The Canadian consumer credit market also faces tight lending standards. The subprime loans

have been a very small segment of the market, eﬀectively reducing delinquencies on auto loans and

credit cards, for example. Moreover, the prevalent use of HELOCs as consumer credit means that

this credit is secured by properties, lowering the overall delinquency rate.

In both episodes, we ﬁnd that the average credit score increased upon the reset across all

contract types. This is consistent with our ﬁndings that (i) in the expansionary episode, borrowers

who reset their rates paid down mortgage debt, and (ii) in the contractionary episode, borrowers

who reset their rates paid down revolving debt.

6 Aggregate Eﬀects of Mortgage Rate Resets

Given our micro-level analysis, we expect mortgage rate resets to generate sizeable eﬀects on

aggregate spending when borrowers reset their rates to lower levels, and to contribute to aggregate

savings when borrowers experience rate increases. Since our credit panel data are representative of

Canadian borrowers, we are able to estimate the aggregate eﬀects on spending and savings.

Speciﬁcally, the eﬀect of rate resets on aggregate spending or savings at time t can be computed

by integrating the corresponding eﬀects across borrowers with diﬀerent types of FRMs who renew

their mortgages at t,

∆R

× 

× φ

(D), (6)

where D denotes the pre-renewal mortgage term. ∆R

is the average change in the mortgage rate

upon the reset for borrowers with term-D FRMs. 

is the interest rate semi-elasticity, i.e., the

change in spending or savings for a 100-bp change in the mortgage rate. φ

(D) is the total number

of borrowers who renew their term-D FRMs at t. We next describe how this formula is computed

for each category: (i) auto spending, (ii) durable consumption, (iii) mortgage pay-down, and (iv)

revolving debt pay-down. A summary of these aggregate eﬀects can be found in Table 12.

Aggregate eﬀect on auto spending. We estimate that the total increase in auto spending

caused by rate resets in the expansionary episode (2015m1-2017m1) was $2.02 billion, equivalent

to 1.53% of the Canadian aggregate new auto sales over this period. This spending estimate is

obtained as follows. ∆R

is set to the corresponding rate change in Table 3 (column 1, panel I).

The interest rate semi-elasticity of auto spending, 

, is set to $1,380 for all borrowers.

(D) is

estimated using Census data on the total number of mortgages in Canada, multiplied by the share

of term-D FRMs, further multiplied by the fraction of renewers among term-D FRM borrowers.

We estimate this eﬀect for 2-, 3-, 4- and 5-year FRM borrowers, add these eﬀects together, and

divide this sum by 95% to reﬂect the share of these four products in the Canadian FRM market.

We also estimate a lower bound for the aggregate auto-spending eﬀect, which is solely based on

the eﬀect of 5-year FRM renewals, setting the eﬀects of all other renewals to zero. This gives an

increase in aggregate auto spending of $1.55 billion, or 1.17% of Canadian aggregate new auto sales

over this period. For the contractionary episode, since we do not ﬁnd a signiﬁcant causal change

in auto spending at the micro level, the estimated aggregate eﬀect is zero.

Aggregate eﬀect on durable consumption. A similar calculation is used for the eﬀect on

aggregate durable consumption. The only diﬀerence is the calibration of 

, which is set to $942.

We estimate that the total increase in durable consumption caused by rate resets in the expansionary

episode (2015m1-2017m1) was $1.38 billion, equivalent to 0.42% of Canadian aggregate durable

expenditures over this period. The aggregate eﬀect of 5-year FRM renewals, which may be viewed

as the lower bound for this eﬀect, was $1.06 billion (0.32% of aggregate durable expenditures). For

the contractionary episode, the aggregate eﬀect was zero, given that no causal evidence was found

for a change in durable expenditures ﬁnanced by installment loans at the micro level.

Aggregate eﬀect on mortgage pay-down. As shown in Section 4, in the expansionary episode,

borrowers used part of their monthly interest savings to pay down their mortgage faster. Here we

assess the total mortgage principal paid down over the lengths of borrowers’ new contracts upon the

resets. For each term D, we multiply the diﬀerence between the required and scheduled payments

(Table 3) by the length of the new term (assumed to be D). The aggregation is similar to that for

estimating aggregate spending. We estimate that $3.56 billion of mortgage debt was paid down in

borrowers’ new terms due to the resets in the expansionary episode. Most of this eﬀect came from

the contribution of 5-year FRM borrowers, who paid down $2.71 billion of mortgage debt in their

This semi-elasticity is computed using the estimated monthly increase in auto spending of 5-year FRM borrowers

($18.56) multiplied by 84 to reﬂect the length of auto loans (usually 7 years), further divided by the rate change of

5-year FRM borrowers upon the reset (113 bps) for normalization.

This semi-elasticity is computed using the estimated monthly increase in spending ﬁnanced by installment loans

of 5-year FRM borrowers ($44.36) multiplied by 24 to reﬂect non-auto durable consumption in 2015m1-2017m1,

further divided by the rate change of 5-year FRM borrowers upon the reset (113 bps) for normalization.

new 5-year term. The aggregate mortgage principal paid down due to resets in the contractionary

episode was much smaller, about $0.4 billion at most.

Aggregate eﬀect on revolving debt pay-down. We apply formula (6) to estimate this eﬀect.

∆R

and φ

(D) are calibrated similarly as before. 

is term-speciﬁc, set to the corresponding

change in total revolving debt (Table 6, column 5) divided by the change in the mortgage rate

(Table 3, column 1) for normalization. The aggregation shows that, in the contractionary episode,

$693 million of revolving debt was paid down, accounting for 1.72% of aggregate saving over this

period. The contribution from 5-year FRM renewing borrowers was relatively small, about $300

million, as deleveraging was across the board (except for 4-year FRM renewers). Since we do not

ﬁnd signiﬁcant change in revolving debt upon the reset in the expansionary episode, the aggregate

eﬀect was zero for that episode.

Time variation in aggregate eﬀects and policy implications. Formula (6) suggests that the

aggregate eﬀect of rate resets may vary over time due to the change in the rate upon the reset, ∆R

even though the interest rate semi-elasticity, 

, and the distribution of borrower types, φ

(D), are

stable over time. The Canadian interest rates have been declining since the global ﬁnancial crisis,

so there were other periods when borrowers experienced sizeable rate changes upon the reset. It

is useful to understand the eﬀects of rate rates in the historical context. To this end, we extend

our analysis from the two recent episodes to the entire period for which the credit panel data are

available and apply formula (6) at the monthly frequency.

Panel (a) of Figure 12 shows the change in the mortgage rate for 5-year FRM borrowers upon

the reset, which displays substantial time variation. The largest decline occurred around December

2012, about 250 bps. These borrowers previously reset their rates in late 2007, right before the

Bank of Canada slashed the policy rate in response to the global ﬁnancial crisis. This sharp decline

in the mortgage rate led to substantial aggregate spending eﬀects, as shown in panels (b) and (c).

For example, total spending on automobiles increased by about $180 million in December 2012,

accounting for almost 6% of the Canadian aggregate new auto sales in that month. Similar patterns

are observed for durable consumption.

These patterns are consistent with those in the related studies: Mortgage borrowers resetting

their rates after the global ﬁnancial crisis through adjustable-rate mortgages in the U.S. (Di Maggio

et al. (2017)) or through variable-rate mortgages in Australia (La Cava et al. (2016)) experienced

substantial consumption growth. Our analysis also suggests that, as countries have been staying

in the low-policy-rate environment for a long time since the ﬁnancial crisis, and as the room for

cutting the conventional policy rate becomes constrained, the reduction in the mortgage rate for

borrowers who experience resets may be limited, weakening the potency of the rate-reset channel

of monetary policy (Berger et al. (2021)).

7 Further Evidence and Robustness Analysis

This section provides additional analysis and robustness checks that support our ﬁndings in Sections

4 and 5. They help address a number of potential concerns arising from the switching of terms,

issues with the baseline identiﬁcation strategy, the anticipation eﬀects, and institutional features

of the Canadian mortgage market.

Switching of terms. As discussed in Section 2.1, some borrowers switch to a diﬀerent term length

during the renewal, while others stay with the same length as in the previous contract. For the

baseline estimates, we do not restrict the term length to be the same before and after the reset.

Thus, these estimates capture the average eﬀects across all post-renewal terms. One concern is that

if borrowers with shorter-term FRMs systematically switch to longer-term FRMs at the renewal or

vice versa, our baseline estimates of rate changes, for example, would be confounded by the rate

spreads between diﬀerent types of mortgages. As shown in Table 2, the mortgage rate is increasing

in the mortgage term.

To address this concern, we redo the analysis in Sections 4 and 5 with the sample that restricts

the term length to be the same before and after the reset. Table D5 shows the results. As expected,

the rate reduction for 2-year FRM borrowers is larger than the baseline in the expansionary episode,

and the rate increase is smaller in the contractionary episode. Changes in rates for other FRM

borrowers are similar to the baseline estimates. The responses of payment choices, consumption

and debt repayment are similar to the baseline estimates. For example, borrowers use part of

their interest savings to pay down mortgage debt in the expansionary episode. In addition, 5-year

FRM borrowers increase their spending when rates decline. Spending does not change when rates

increase. We also see that borrowers deleverage on revolving debt in the contractionary episode

and ﬁnd no evidence of rising delinquencies.

Diﬀerence-in-diﬀerence (DID) estimates. As discussed in Section 3, one concern with our

baseline strategy is the inability to account for the mortgage-age eﬀect that could confound the

eﬀect of rate resets. To address this concern, we employ a DID design with the control group

being 7- and 10-year FRMs previously reset at the same time as the FRMs in our baseline sample.

Table D6 shows the results using this approach. At the loan level, it conﬁrms that mortgages reset

in the two episodes experienced substantial changes in rates and payments with the magnitudes

varying with the pre-renewal term. The asymmetric mortgage prepayment choices are preserved

as in Section 4.2. The consumer-level responses of spending, debt repayment and delinquencies

are quantitatively similar to the baseline estimates. The only noticeable diﬀerence is that, in the

expansionary episode, deleveraging appears on both credit card debt and LOC debt.

The main reason why we do not use this alternative speciﬁcation as the baseline strategy is that

the size of the control group is small, as not many Canadian borrowers take longer-term FRMs. In

the data, 7- and 10-year FRMs account for only 2% of the mortgage stock. In addition, one might

be concerned about the endogenous selection into these mortgage products. For these reasons, we

estimate a second DID speciﬁcation with the control group being FRMs that had the same terms as

the treatment group but were not renewed in the two episodes. As shown in Table D7, these DID

estimates are similar to the baseline estimates and to the ﬁrst set of DID estimates. One caveat of

the second DID approach is that we are unable to obtain a control group for 2-year FRMs, because

all existing 2-year FRMs would be renewed in a 2-year episode.

Anticipation eﬀects. An interesting question is whether the consumer-level responses in Section

5 should be interpreted as the responses to anticipated mortgage rate changes. Economic theory

suggests that unconstrained, rational-expectations consumers respond to the news of the shock, not

the anticipated realizations of the shock. It is possible that consumers may have already anticipated

the change in their mortgage rates before the actual reset. If this is true and if consumers are not

liquidity constrained, in the expansionary episode, we would see a jump in their spending before

the actual reset. This is not what we see in the data, however. Spending increased only when

consumers actually reset their rates. One may argue that this is due to the presence of the liquidity

constraint that prevents consumers from reacting before the cash ﬂow arrives. This is unlikely,

because liquidity-constrained consumers only account for a small fraction of the population.

Another way of assessing the importance of the anticipation eﬀect is to compare consumers’

responses in a period when mortgage rate changes are unexpected to the responses in a period

when rate changes are of similar sizes but more likely to be anticipated. The monetary policy rate

cut in January 2015 was widely considered as a surprise to the market.

We therefore compare the

responses of borrowers who renewed their mortgages in 2015Q1 to the responses of borrowers who

renewed their mortgages in 2015Q2. Our premise is that, if the anticipation eﬀect is important,

borrowers in the former group should have stronger responses than in the latter group. We do

not ﬁnd signiﬁcantly diﬀerent responses across the two groups of borrowers. This pattern holds

even when we compare borrowers who who renewed their mortgages in February 2015 to those who

renewed their mortgages in March 2015. These results suggest that anticipation eﬀects, if they

exist, are weak and are unlikely to alter our estimates.

Ahead-of-schedule renewals. As shown in Figure 2, about 50% of borrowers renew their

mortgages earlier than scheduled. This raises the question of whether on-time renewals and

ahead-of-schedule renewals result in diﬀerent responses to the same shock. We perform additional

analysis to address this question. First, we restrict the sample to borrowers who renew their

contracts on time and estimate their responses to rate resets. The results are similar to the

baseline estimates. Second, we interact the post-renewal indicator with a set of dummies that

indicate the months ahead of the scheduled renewal. Again, we do not ﬁnd diﬀerent responses of

borrowers making ahead-of-schedule renewals. We do ﬁnd, however, that the mortgage rates from

early renewals are slightly lower than on-time renewals, but the diﬀerences are small compared to

the overall rate changes.

Mortgages renewed in both episodes. When assessing the asymmetric eﬀects of rate resets,

we would ideally focus on the same borrowers who experienced resets in both episodes. Due to

the timing of the two episodes, we only observe a subset of 2- and 3-year FRM borrowers who

experienced resets in both episodes. None of the 4- or 5-year FRM borrowers in our sample had

two resets with one in each episode. Nevertheless, we perform a robustness check using the sample

of the 2- and 3-year FRM borrowers who experience both resets. In unreported results, we show

that the loan-level and consumer-level responses are very similar to those in Table D5.

Although the monetary policy statement makes it explicit that the decision was in response to the sharp drop

in oil prices, the decision, when it came, was unexpected by many observers. In fact, the market had been predicting

a rate increase later that year. See, for example, Shecter (2015), “Bank of Canada’s surprise rate cut seen hurting

Canadian banks’ proﬁts,” Financial Post, January 21, 2015; “Bank of Canada shocks markets with cut in key interest

rate,” CBC Business News, January 21, 2015.

This ﬁnding may be explained by consumers’ inattention to movements in interest rates or the uncertainty about

the realizations of the mortgage rate change.

8 Conclusion

One of the most important channels through which monetary policy aﬀects the real economy is

changes in mortgage rates. This channel is particularly relevant for policymakers in countries

dominated by variable-rate mortgages, adjustable-rate mortgages and Canadian-type ﬁxed-rate

mortgages, as changes in the policy rate in these countries are passed through automatically to

most homeowners, aﬀecting their consumption and savings.

We study the eﬀects of mortgage rate changes driven by monetary policy shifts on consumer

spending, debt repayment, and defaults in Canada, taking advantage of the institutional features of

the Canadian mortgage market. This setting facilitates the design of a clean identiﬁcation strategy

for causal inference. In addition, the high-quality consumer credit panel data allow us to examine

changes in a consumer’s entire credit portfolio at a monthly frequency. Most importantly, we are

able to provide a detailed analysis of how consumers respond to rate increases and decreases.

Our ﬁndings for the expansionary episode, while new, are broadly in line with those using data

from other countries. Consumers increase durable spending, pay down mortgage debt, and reduce

the likelihood of being delinquent. In the cross-section, liquidity-constrained borrowers obtain

more cash from their interest savings, but their ability to use debt to ﬁnance durable spending is

limited by their access to new credit. Since cash ﬂows resulting from lower mortgage payments are

realized over the course of several years, the diﬃculty in accessing credit markets may dampen the

immediate eﬀect of monetary stimulus on consumer spending.

Our ﬁndings for the contractionary episode call into question the conventional wisdom.

Speciﬁcally, we do not ﬁnd that durable spending decreases when mortgage rates increase. This

implies that consumers either dissave to maintain their consumption or cut other types of spending.

We also document a robust pattern that consumers lower, rather than increase, their revolving

debt level, which cannot be reconciled with the cash-ﬂow interpretation of rate resets. We provide

evidence that this pattern is explained by changes in expectations about future rates. Finally, we

do not see rising delinquencies or tightening of credit supply for borrowers who reset their rates to

higher levels.

Our analysis of the contractionary episode suggests that mortgage rate resets do not appear to

discourage durable spending, render consumers more leveraged or increase the likelihood of defaults,

as commonly asserted in the ﬁnancial press. Of course, our paper examines only one aspect of

mortgage rate changes, namely, the resets experienced by existing homeowners. There are other

channels through which mortgage rate changes may aﬀect households, for example, home sales and

house prices, wealth eﬀects resulting from changing asset prices, and home equity extraction. We

leave these issues for future research.

The Canadian-type short-term FRMs we focus on are quite common in other OECD countries.

Previous studies of the relationship between mortgage payments and consumer behavior have largely

relied on data from the U.S.. It is unclear whether the U.S. evidence can be generalized to other

countries. The U.S. is unique in having an unusually high share of long-term FRMs, the extensive

use of securitization in housing ﬁnance, and the absence of prepayment penalties (see Lea (2010)).

Although our estimates may depend on the choice of the episodes and the Canadian socio-economic

conditions, our qualitative insights should apply more broadly to other countries.

References

Abel, J. and A. Fuster, “How Do Mortgage Reﬁnances Aﬀect Debt, Default, and Spending?

Evidence from HARP,” American Economic Journal: Macroeconomics 13(2) (2021), 254–291,

https://doi.org/10.1257/mac.20180116.

Agarwal, S., G. Amromin, I. Ben-David, S. Chomsisengphet, T. Piskorski

and A. Seru, “Policy Intervention in Debt Renegotiation: Evidence from Home

Aﬀordability Modiﬁcation Program,” Journal of Political Economy 125(3) (2017), 654–712,

https://doi.org/10.1086/691701.

Agarwal, S., Y. Deng, Q. Gu, J. He, W. Qian and Y. Ren, “Mortgage Debt, Hand-to-Mouth

Households, and Monetary Policy Transmission,” forthcoming: Review of Finance (2022),

https://doi.org/10.1093/rof/rfac011.

Agarwal, S., C. Liu and N. S. Souleles, “Reaction of Consumer Spending and Debt to Tax

Rebates: Evidence from Consumer Credit Data,” Journal of Political Economy 115(6) (2007),

986–1019, https://doi.org/10.1086/528721.

Agarwal, S. and W. Qian, “Consumption and Debt Response to Unanticipated Income Shocks:

Evidence from a Natural Experiment in Singapore,” American Economic Review 104(12) (2014),

4205–4230, https://doi.org/10.1257/aer.104.12.4205.

Agarwal, S., R. Rosen and V. Yao, “Why Do Borrowers Make Mortgage

Reﬁnancing Mistakes?,” Management Science 62(12) (2016), 3494–3509,

https://doi.org/10.1287/mnsc.2015.2272.

Allen, J., R. Clark and J. Houde, “Price Dispersion in Mortgage Markets,” Journal of

Industrial Economics 62(3) (2014), 377–416, https://doi.org/10.1111/joie.12046.

———, “Search Frictions and Market Power in Negotiated Price Markets,” Journal of Political

Economy 127(4) (2019), 1550–1598, https://doi.org/10.1086/701684.

Andersen, S., J. Y. Campbell, K. M. Nielsen and T. Ramadorai, “Sources of Inaction in

Household Finance: Evidence from the Danish Mortgage Market,” American Economics Review

110(10) (2020), 3184–3230, https://doi.org/10.1257/aer.20180865.

Bailliu, J., K. Kartashova and C. Meh, “Household Borrowing and Spending in Canada,”

Bank of Canada Review, Winter, 2012.

Baugh, B., I. Ben-David, H. Park and J. A. Parker, “Asymmetric Consumption Smoothing,”

American Economic Review 111(1) (2021), 192–230, https://doi.org/10.1257/aer.20181735.

Beraja, M., A. Fuster, E. Hurst and J. Vavra, “Regional Heterogeneity and the

Reﬁnancing Channel of Monetary Policy,” Quarterly Journal of Economics 134(1) (2019),

109–183, https://doi.org/10.1093/qje/qjy021.

Berger, D., K. Milbradt, F. Tourre and J. Vavra, “Mortgage Prepayment and

Path-Dependent Eﬀects of Monetary Policy,” American Economic Review 111(9) (2021),

2829–2878, https://doi.org/10.1257/aer.20181857.

Bhutta, N. and B. J. Keys, “Interest Rates and Equity Extraction during the Housing Boom,”

American Economic Review 106(7) (2016), 1742–1774, https://doi.org/10.1257/aer.20140040.

Carroll, C. D., “Buﬀer-Stock Saving and the Life Cycle/Permanent Income Hypothesis,”

Quarterly Journal of Economics 112(1) (1997), 1–55, https://doi.org/10.1162/003355397555109.

Chen, H., M. Michaux and N. Roussanov, “Houses as ATMs? Mortgage

Reﬁnancing and Macroeconomic Uncertainty,” Journal of Finance 75(1) (2020), 323–375,

https://doi.org/10.1111/joﬁ.12842.

Cloyne, J., C. Ferreira and P. Surico, “Monetary Policy when Households have Debt: New

Evidence on the Transmission Mechanism,” Review of Economic Studies 87(1) (2020), 102–129,

https://doi.org/10.1093/restud/rdy074.

Crawford, A., C. Meh and J. Zhou, “The Residential Mortgage Market in Canada: A Primer,”

Bank of Canada Financial System Review, 2013.

Deaton, A., “Saving and Liquidity Constraints,” Econometrica 59(5) (1991), 1221–1248,

https://doi.org/10.2307/2938366.

DeFusco, A. A. and J. Mondragon, “No Job, No Money, No Reﬁ: Frictions to Reﬁnancing in

a Recession,” Journal of Finance 75(5) (2020), 2327–2376, https://doi.org/10.1111/joﬁ.12952.

Di Maggio, M., B. Keys, T. Piskorski, R. Ramcharan, A. Kermani, A. Seru

and V. Yao, “Interest Rate Pass-Through: Mortgage Rates, Household Consumption,

and Voluntary Deleveraging,” American Economic Review 107 (2017), 3350–3388,

https://doi.org/10.1257/aer.20141313.

Ehrlich, G. and J. Perry, “Do Large-Scale Reﬁnancing Programs Reduce Mortgage Defaults?

Evidence from a Regression Discontinuity Design,” University of Michigan Working Paper, 2017.

Floden, M., M. Kilstrom, J. Sigurdsson and R. Vestman, “Household Debt and Monetary

Policy: Revealing the Cash-ﬂow Channel,” The Economic Journal 131(636) (2021), 1742–1771,

https://doi.org/10.1093/ej/ueaa135.

Fuster, A. and P. Willen, “Payment Size, Negative Equity, and Mortgage

Default,” American Economic Journal: Economic Policy 9(4) (2017), 167–191,

https://doi.org/10.1257/pol.20150007.

Ganong, P. and P. Noel, “Liquidity vs. Wealth in Household Debt Obligations: Evidence from

Housing Policy in the Great Recession,” American Economic Review 110(10) (2020), 3100–3138,

https://doi.org/10.1257/aer.20181243.

Gosselin, M.-A. and M. Khan, “A Survey of Consumer Expectations for Canada,” Bank of

Canada Review, Autumn, 2015.

Greenwald, D. L., “The Mortgage Credit Channel of Macroeconomic Transmission,” MIT Sloan

Research Paper No. 5184-16, 2017.

Ho, A. T. Y., M. Khan, M. Mow and B. Peterson, “Home Equity Extraction and Household

Spending in Canada,” Staﬀ Analytical Note, 2019.

Hurst, E. and F. P. Stafford, “Home Is Where the Equity Is: Mortgage Reﬁnancing and

Household Consumption,” Journal of Money, Credit, and Banking 36(6) (2004), 985–1014,

https://doi.org/10.1353/mcb.2005.0009.

Jappelli, T. and A. Scognamiglio, “Interest Rate Changes, Mortgages, and

Consumption: Evidence from Italy,” Economic Policy 33 (2018), 183–224,

https://doi.org/10.1093/epolic/eiy001.

Johnson, D., J. A. Parker and N. S. Souleles, “Household Expenditure and the

Income Tax Rebates of 2001,” American Economic Review 96(5) (2006), 1589–1610,

https://doi.org/10.1257/aer.96.5.1589.

Kaplan, G., B. Moll and G. L. Violante, “Monetary Policy According to HANK,” American

Economic Review 108(3) (2018), 697–743, https://doi.org/10.1257/aer.20160042.

Kaplan, G. and G. L. Violante, “A Model of the Consumption Response to Fiscal Stimulus

Payments,” Econometrica 82(4) (2014), 1199–1239, https://doi.org/10.3982/ECTA10528.

Karamon, K., D. McManus and J. Zhu, “Reﬁnance and Mortgage Default: A Regression

Discontinuity Analysis of HARP’s Impact on Default Rates,” Journal of Real Estate Finance

and Economics 55(4) (2017), 457–475, https://doi.org/10.1007/s11146-016-9566-z.

Keys, B. J., D. G.Pope and J. C. Pope, “Failure to Reﬁnance,” Journal of Financial Economics

122(3) (2016), 482–499, https://doi.org/10.1016/j.jﬁneco.2016.01.031.

Kilian, L. and X. Zhou, “The Propagation of Regional Shocks in Housing Markets: Evidence

from Oil Price Shocks in Canada,” forthcoming: Journal of Money, Credit and Banking (2020),

https://doi.org/10.1111/jmcb.12847.

La Cava, G., H. Hughson and G. Kaplan, “The Household Cash Flow Channel of Monetary

Policy,” Reserve Bank of Australia Discussion Paper 2016-12, 2016.

Lea, M., “International Comparison of Mortgage Product Oﬀerings,” Special Report. Research

Institute for Housing America, 2010.

Parker, J. A., N. S. Souleles, D. S. Johnson and R. McClelland, “Consumer Spending

and the Economic Stimulus Payments of 2008,” American Economic Review 103(3) (2013),

2530–2553, https://doi.org/10.1257/aer.103.6.2530.

Tracy, J. and J. Wright, “Payment Changes and Default Risk: The Impact of

Reﬁnancing on Expected Credit Losses,” Journal of Urban Economics 93 (2016), 60–70,

https://doi.org/10.1016/j.jue.2016.03.007.

Watts, A., “Canada’s Evolving Car Sales Market: Evidence from J.D. Power Auto Data,” Bank

of Canada Internal Note, August 16, 2016.

Wong, A., “Reﬁnancing and the Transmission of Monetary Policy to Consumption,” Princeton

University Working Paper, 2021.

Zhou, X., “Mortgage Borrowing and the Boom-Bust Cycle in Consumption and

Residential Investment,” Review of Economic Dynamics 44 (2022), 244–268,

https://doi.org/10.1016/j.red.2021.03.004.

Figure 1: Bank of Canada policy rate (overnight target rate)

0 1 2 3 4 5

percent

2007m1 2009m1 2011m1 2013m1 2015m1 2017m1 2019m1

Source: Bank of Canada. The overnight rate is the interest rate at which major ﬁnancial institutions borrow and

lend one-day (or “overnight”) funds among themselves. The Bank of Canada sets a target level for that rate, often

referred to as the Bank’s policy rate. The ﬁrst vertical line indicates the beginning of our micro data. The other two

lines indicate the beginning of the two episodes in our study, 2015m1 and 2017m7.

Figure 2: Distribution of the timing of mortgage renewal

0 10 20 30 40 50

percent

−6 −5 −4 −3 −2 −1 0 1 2 3 4 5 6

Months since scheduled renewal

Notes: This ﬁgure plots the fraction of borrowers who renew their mortgages x months after the scheduled renewal,

where x = 0 refers to on-time renewal.

Figure 3: Spending responses of 5-year FRM borrowers around the reset in the expansionary episode

-2 0 2 4

quarter

-20

dollars/month

Auto spending

-2 0 2 4

quarter

-200

200

400

600

dollars

Cumulative auto spending

-2 0 2 4

quarter

-50

100

dollars/month

IL-Financed spending

-2 0 2 4

quarter

-500

500

1000

dollars

Cumulative IL-financed spending

Notes: Point estimates and 95% conﬁdence intervals obtained by estimating equation (2) for 5-year FRM borrowers

in the expansionary episode. The left column shows the dynamic responses of monthly auto spending and monthly

spending ﬁnanced by installment loans (IL) to rate resets. The right column shows the dynamic responses of

cumulative spending.

Figure 4: Responses of revolving debt balances around the reset

(a) Expansionary episode

-2 0 2 4

quarter

-600

-400

-200

200

dollars

Credit card (FRM-2yr)

-2 0 2 4

quarter

-1000

1000

2000

3000

dollars

Lines of credit (FRM-2yr)

-2 0 2 4

quarter

-400

-200

200

dollars

Credit card (FRM-5yr)

-2 0 2 4

quarter

-500

500

1000

1500

dollars

Lines of credit (FRM-5yr)

(b) Contractionary episode

-2 0 2 4

quarter

-600

-400

-200

200

dollars

Credit card (FRM-2yr)

-2 0 2 4

quarter

-2000

-1000

1000

dollars

Lines of credit (FRM-2yr)

-2 0 2 4

quarter

-600

-400

-200

200

400

dollars

Credit card (FRM-5yr)

-2 0 2 4

quarter

-2000

-1000

1000

2000

dollars

Lines of credit (FRM-5yr)

Notes: Point estimates and 95% conﬁdence intervals obtained by estimating equation (2) for 2-year and 5-year FRM

borrowers in the expansionary (panel a) and contractionary episode (panel b).

Figure 5: Mortgage rate resets and aggregate spending eﬀects, 2009-1019

(a) Average change in the mortgage rate upon the reset, 5-year FRM borrowers

−3 −2 −1 0 1

percent

2009m1 2011m1 2013m1 2015m1 2017m1 2019m1

(b) Eﬀect of rate resets on aggregate auto spending

0 50 100 150 200

million $

2009m1 2011m1 2013m1 2015m1 2017m1 2019m1

All FRM renewing borrowers

5−year FRM renewing borrowers

0 50 100 150

million $

2009m1 2011m1 2013m1 2015m1 2017m1 2019m1

All FRM renewing borrowers

5−year FRM renewing borrowers

Notes: Panel (a) depicts the average change in the mortgage rate experienced by 5-year FRM borrowers who

renewed their contracts in a given month. Panels (b) and (c) plot the estimated aggregate spending eﬀects using the

methodology described in Section 6.

Table 1: Mortgage characteristics at origination

Bank of the sample Other lenders

Mean Median Mean Median

All FRMs

Market share (%) 18 - 82 -

Contract rate (%) 2.89 2.84 2.90 2.79

Outstanding balance ($) 289,766 248,541 302,050 255,745

LTV ratio (%) 78.6 80.0 77.9 80.0

DTI ratio (%) 329.0 302.1 334.8 296.8

Credit score 768 771 756 763

Borrower age 42.5 41.0 41.9 40.0

Fraction of insured mortgages (%) 33.1 - 35.9 -

Fraction of FRM-5yr (%) 64.1 - 58.0 -

FRM-5yr

Market share (%) 19 - 81 -

Contract rate (%) 2.90 2.82 2.88 2.79

Outstanding balance ($) 307,691 266,540 291,600 255,272

LTV ratio (%) 80.0 80.5 80.7 80.0

DTI ratio (%) 352.3 332.3 340.7 313.9

Credit score 765 768 756 762

Borrower age 41.4 39.0 41.0 39.0

Fraction of insured mortgages (%) 38.5 - 45.4 -

Source: Bank of Canada-OSFI mortgage originations dataset. This table shows the characteristics of the mortgages

originated by the bank of our sample and by all other federally regulated lenders between 2014 and 2018 for the

purpose of home purchases.

Table 2: Summary statistics

FRM-2yr FRM-3yr FRM-4yr FRM-5yr

Mean SD Mean SD Mean SD Mean SD

Panel I: FRMs renewed in 2015m1-2017m1

Current balance ($) 160,192 131,466 166,744 128,491 181,191 125,649 169,687 119,397

Mortgage rate (%) 2.55 0.34 2.66 0.33 2.79 0.34 3.58 0.69

Scheduled payment ($/month) 950 626 987 610 1,070 604 995 585

Number of loans 23,023 17,105 7,251 40,949

Corresponding consumers

Age 51.31 12.82 50.51 12.83 49.70 12.68 48.34 13.00

Credit score 770 103 767 104 776 100 749 113

Credit utilization rate 0.37 0.35 0.37 0.35 0.34 0.33 0.44 0.36

Auto spending ($/month) 121 2,186 123 2,179 127 2,227 114 2,078

Prob. of an auto purchase (%) 0.38 6.15 0.40 6.28 0.41 6.37 0.38 6.15

Installment loan-ﬁnanced spending ($/month) 237 3,797 222 3,650 212 3,320 240 3,486

Prob. of an installment loan origination (%) 0.86 9.25 0.83 9.09 0.80 8.90 0.97 9.79

Credit card balance ($) 4,195 6,910 4,279 7,001 4,117 6,849 4,215 6,831

Lines of credit balance ($) 19,179 38,836 18,559 37,841 17,612 36,238 14,601 29,649

60-day mortgage delinquency rate (%) 0.74 27.2 0.90 30.0 0.83 28.3 1.87 43.2

60-day auto loan delinquency rate (%) 0.15 12.1 0.09 9.5 0.08 8.9 0.16 12.5

60-day installment loan delinquency rate (%) 0.32 18.0 0.36 19.0 0.30 17.2 0.66 25.8

60-day credit card delinquency rate (%) 3.23 56.7 3.52 59.3 2.87 53.5 4.70 68.4

60-day lines of credit delinquency rate (%) 0.84 28.9 0.98 31.2 0.83 28.8 1.12 33.4

Panel II: FRMs renewed in 2017m7-2019m6

Current balance ($) 195,534 209,237 162,626 117,453 155,373 104,203 199,594 132,347

Mortgage rate (%) 2.51 0.49 2.66 0.42 2.83 0.32 3.12 0.38

Scheduled payment ($/month) 1,059 938 968 597 947 529 1,141 642

Number of loans 30,606 7,056 16,476 31,238

Corresponding consumers

Age 51.91 13.17 51.79 12.95 51.64 14.41 49.14 13.10

Credit score 767 106 761 110 766 106 759 110

Credit utilization rate 0.37 0.35 0.39 0.35 0.40 0.35 0.40 0.35

Auto spending ($/month) 152 2,592 131 2,352 122 2,335 121 2,268

Prob. of an auto purchase (%) 0.44 6.65 0.39 6.26 0.36 6.01 0.37 6.07

Installment loan-ﬁnanced spending ($/month) 300 5,775 265 4,128 277 4,116 277 4,442

Prob. of an installment loan origination (%) 0.99 9.89 0.95 9.70 0.98 9.87 1.01 10.02

Credit card balance ($) 4,587 7,333 4,593 7,281 4,416 7,154 4,799 7,484

Lines of credit balance ($) 20,802 44,112 19,377 40,181 18,863 37,361 18,045 39,517

60-day mortgage delinquency rate (%) 0.59 24.3 0.91 30.2 0.77 27.7 1.29 36.0

60-day auto loan delinquency rate (%) 0.16 12.5 0.14 11.8 0.13 11.2 0.16 12.7

60-day installment loan delinquency rate (%) 0.46 21.4 0.49 22.1 0.48 21.9 0.54 23.3

60-day credit card delinquency rate (%) 3.16 56.1 3.26 57.0 3.42 58.3 4.16 64.4

60-day lines of credit delinquency rate (%) 0.63 25.0 0.73 27.0 0.76 27.5 0.82 28.7

Source: TransUnion Canada account-level data.

Table 3: Mortgage loan-level adjustments upon the reset

Mortgage rate Required payment Scheduled payment Amortization

(p.p.) ($/month) ($/month) (months)

(1) (2) (3) (4)

Panel I: Expansionary episode

FRM-5yr

PostRenew -1.13*** -92.03*** -46.47*** -13.97***

(0.004) (0.55) (0.64) (0.20)

FRM-4yr

PostRenew -0.38*** -34.17*** -9.90*** -6.05***

(0.007) (0.82) (1.85) (0.36)

FRM-3yr

PostRenew -0.18*** -13.91*** -2.19 -4.44***

(0.004) (0.51) (1.17) (0.21)

FRM-2yr

PostRenew -0.16*** -14.74*** -1.76** -4.87***

(0.003) (0.38) (0.88) (0.18)

Panel II: Contractionary episode

FRM-5yr

PostRenew 0.32*** 34.00*** 39.23*** -1.64***

(0.003) (0.45) (0.73) (0.11)

FRM-4yr

PostRenew 0.49*** 36.29*** 40.37*** -1.09***

(0.003) (0.34) (0.77) (0.13)

FRM-3yr

PostRenew 0.70*** 54.98*** 49.49*** 0.66***

(0.006) (0.77) (1.31) (0.24)

FRM-2yr

PostRenew 0.85*** 83.33*** 84.49*** -1.38***

(0.003) (0.66) (0.81) (0.12)

Notes: Each cell presents the results from estimating one regression using equation (1). ** and *** denote signiﬁcance

levels at 5% and 1%. Standard errors are clustered at the loan level. All regressions include a set of control variables

(see Section 3), month ﬁxed eﬀects, and loan ﬁxed eﬀects.

Table 4: Estimated cash ﬂows from the rate reset

Panel I: Expansionary episode Panel II: Contractionary episode

Remaining Interest Amortization-adjusted Remaining Interest Amortization-adjusted

months savings ($) interest savings ($) months savings ($) interest savings ($)

FRM-5yr 227 +20,891 +23,925 208 -7,072 -6,242

FRM-4yr 208 +7,107 +8,485 197 -7,149 -6,889

FRM-3yr 209 +2,907 +4,830 205 -11,271 -11,904

FRM-2yr 219 +3,228 +4,998 230 -19,165 -17,880

Notes: The ﬁrst column in each panel shows the remaining time for paying oﬀ the mortgage before the reset. The

second column in each panel shows the unadjusted total interest savings from the rate reset (obtained by multiplying

the change in the required monthly payment by the remaining months). The third column in each panel shows the

total interest savings adjusted for the change in the amortization.

Table 5: Heterogeneity in the mortgage payment choice (expansionary episode)

Required Scheduled Realization Required Scheduled Realization Required Scheduled Realization

payment payment rate (%) payment payment rate (%) payment payment rate (%)

FRM-5yr

PostRenew -82.24*** -30.02*** 36.5 -82.11*** -29.94*** 36.5 -83.11*** -39.30*** 47.3

(0.63) (1.07) (0.62) (1.05) (0.67) (1.07)

PostRenew -14.09*** -32.90*** 65.3

×LowScore (0.92) (1.44)

PostRenew -16.18*** -36.53*** 67.6

×HighUse (0.95) (1.46)

PostRenew -23.83*** -21.40*** 56.8

×Young (1.00) (1.55)

PostRenew 22.87*** 3.92** 58.7

×Old (1.15) (1.90)

FRM-4yr

PostRenew -35.18*** -8.33*** 23.7 -34.98*** -7.53*** 21.5 -32.37*** -6.90*** 21.3

(0.77) (2.24) (0.77) (2.16) (0.91) (2.17)

PostRenew 2.60** -4.20 38.5

×LowScore (1.28) (2.77)

PostRenew 3.14** -8.68*** 50.9

×HighUse (1.38) (2.84)

PostRenew -7.09*** -6.92** 35.0

×Young (1.41) (3.00)

PostRenew 5.15*** -7.19 51.8

×Old (1.50) (3.98)

Notes: LowScore refers to borrowers whose average credit score in the previous 12 months is below the median of the

distribution. HighUse refers to borrowers whose average credit utilization rate in the previous 12 months is greater

than 0.5. Young and old borrowers refer to ages below 45 and ages above 65, respectively. The realization rate is

obtained by dividing the change in the scheduled monthly payment by the change in the required monthly payment.

Table 6: Responses of spending and revolving debt balances

Auto spending Auto purchase IL-Financed IL-Financed Tot revolving Credit card LOC HELOC

($/m) prob. (%) spending ($/m) purchase prob. (%) debt ($) debt ($) debt ($) debt ($)

(1) (2) (3) (4) (5) (6) (7) (8)

Panel I: Expansionary episode

FRM-5yr

PostRenew 18.56*** 0.073*** 44.36*** 0.141*** 101.31 -160.90*** 251.98** 449.60***

(6.09) (0.017) (12.03) (0.029) (124.59) (32.54) (120.79) (104.53)

FRM-4yr

PostRenew -21.53 -0.053 19.83 0.111 193.69 -133.00** 247.81 501.49

(19.61) (0.054) (30.63) (0.084) (352.90) (65.79) (353.62) (306.57)

FRM-3yr

PostRenew 13.81 0.036 30.25 0.051 -329.1 -247.60*** 3.22 155.89

(10.29) (0.029) (17.55) (0.043) (191.67) (38.99) (188.11) (164.79)

FRM-2yr

PostRenew -4.24 -0.008 33.88 0.088** 49.02 -167.40*** 246.69** 138.14

(10.30) (0.029) (20.20) (0.044) (124.86) (25.78) (123.34) (112.42)

Panel II: Contractionary episode

FRM-5yr

PostRenew 7.09 0.024 22.36 0.036 -438.20** -246.90*** -278.80 -37.03

(8.34) (0.022) (15.55) (0.035) (213.42) (40.73) (210.50) (186.23)

FRM-4yr

PostRenew 6.34 -0.002 11.77 0.079 167.26 -247.60*** 428.15 193.13

(12.57) (0.033) (23.68) (0.056) (226.53) (48.94) (220.17) (199.38)

FRM-3yr

PostRenew 16.90 0.048 41.46 0.087 -900.70*** -273.90*** -596.50 -238.13

(18.57) (0.049) (28.89) (0.075) (323.86) (64.12) (312.23) (277.38)

FRM-2yr

PostRenew 20.52 0.066** 44.59 0.068 -261.60** -213.30*** -44.50 196.35

(10.73) (0.027) (23.73) (0.040) (133.72) (24.36) (131.97) (121.76)

Notes: Each cell presents the results from estimating one regression using equation (1). ** and *** denote signiﬁcance

levels at 5% and 1%. Standard errors are clustered at the consumer level. All regressions include a set of control

variables (see Section 3), month ﬁxed eﬀects, and consumer ﬁxed eﬀects.

Table 7: Heterogeneity in the spending responses (expansionary episode, FRM-5yr)

Auto spending IL-Financed Auto spending IL-Financed Auto spending IL-Financed

($/m) spending ($/m) ($/m) spending ($/m) ($/m) spending ($/m)

PostRenew 24.69*** 73.24** 17.67*** 31.88** 15.55** 32.95**

(6.64) (14.34) (6.73) (12.81) (6.87) (13.82)

PostRenew -11.21 -49.29***

×LowScore (6.64) (12.73)

PostRenew -1.75 16.39

×HighUse (7.10) (12.85)

PostRenew 7.04 42.72***

×Young (7.69) (14.36)

PostRenew -4.79 -29.16**

×Old (7.85) (14.67)

Notes: LowScore refers to borrowers whose average credit score in the previous 12 months is below the median of the

distribution. HighUse refers to borrowers whose average credit utilization rate in the previous 12 months is greater

than 0.5. Young and old borrowers refer to ages below 45 and ages above 65, respectively.

Table 8: Heterogeneity in the responses of revolving debt balances

Revolving CC LOC Credit Revolving CC LOC Credit Revolving CC LOC Credit

($) ($) ($) Utilization ($) ($) ($) Utilization ($) ($) ($) Utilization

Panel I: Expansionary episode

FRM 5-yr

PostRenew -509.27*** -407.72*** -137.33 -0.033*** -3100*** -566.32*** -2600*** -0.042*** 825.97*** -237.46*** 1110*** -0.021***

(166.40) (37.91) (162.58) (0.002) (155.20) (37.37) (151.30) (0.002) (169.64) (41.48) (164.39) (0.002)

PostRenew 1080*** 409.52*** 724.74*** 0.023***

×LowScore (202.65) (46.63) (197.75) (0.002)

PostRenew 6878*** 759.84*** 6244*** 0.051***

×HighUse (210.24) (51.25) (205.04) (0.002)

PostRenew -1100*** 254.92*** -1500*** 0.000

×Young (207.00) (48.43) (200.84) (0.002)

PostRenew -2100*** -140.09** -2100*** 0.007**

×Old (290.76) (69.82) (289.53) (0.003)

Panel II: Contractionary episode

FRM-2yr

PostRenew -589.50*** -277.23*** -340.14 -0.017*** -389.32** -268.95*** -106.29 -0.004*** 390.15 -185.31*** 580.06*** -0.011***

(202.81) (30.88) (199.91) (0.001) (177.75) (29.21) (175.31) (0.001) (201.19) (34.01) (197.70) (0.001)

PostRenew 490.53 85.09 461.19 0.006***

×LowScore (288.80) (49.93) (282.29) (0.002)

PostRenew 469.92 156.48** 310.75 -0.018***

×HighUse (314.78) (64.17) (305.56) (0.002)

PostRenew -497.17 141.08*** -638.29*** 0.005

×Young (305.52) (53.98) (299.28) (0.002)

PostRenew -2100*** -192.69*** -1900*** -0.006**

×Old (389.55) (64.43) (380.15) (0.003)

Notes: LowScore refers to borrowers whose average credit score in the previous 12 months is below the median of the distribution. HighUse refers to borrowers

whose average credit utilization rate in the previous 12 months is greater than 0.5. Young and old borrowers refer to ages below 45 and ages above 65, respectively.

Credit utilization rate is the ratio of the total revolving balance over the total revolving credit limit.

Table 9: Responses of revolving credit supply (contractionary episode)

Prob. higher Prob. higher CC limit LOC limit CC payment-to- LOC payment-to-

CC limit LOC limit ($) ($) balance ratio balance ratio

(1) (2) (3) (4) (5) (6)

FRM-5yr

PostRenew 0.002 0.010*** -109.70 2553*** 0.007 0.002

(0.001) (0.001) (66.17) (312.88) (0.005) (0.018)

FRM-4yr

PostRenew 0.005*** 0.013*** 68.17 2899*** -0.003 0.030

(0.001) (0.001) (74.91) (329.10) (0.007) (0.018)

FRM-3yr

PostRenew 0.004 0.015*** 114.41 2369*** 0.001 0.001

(0.002) (0.002) (96.23) (486.41) (0.009) (0.030)

FRM-2yr

PostRenew 0.003*** 0.010*** -62.4 2492*** 0.002 -0.014

(0.001) (0.001) (34.54) (181.70) (0.004) (0.023)

Notes: Columns (1)-(2) show changes in the likelihood of increasing the credit limit by more than $1,000. Columns

(3)-(4) show changes in the credit limit. Columns (5)-(6) show changes in the required payment to balance ratio.

Table 10: Evidence from the Canadian Survey of Consumer Expectations

Rates higher Rates higher Rates higher Pay down Cut spending Postpone Bring fwd.

in 1 year in 1&2 years in 1,2 &5 years debt save more purchases purchases

(1) (2) (3) (4) (5) (6) (7)

Panel I: All periods

Rates rising recently 0.217*** 0.215*** 0.166*** 0.096*** 0.078*** 0.029*** -0.009

(0.012) (0.013) (0.012) (0.013) (0.013) (0.011) (0.007)

Controls Y Y Y Y Y Y Y

Quarter ﬁxed eﬀects Y Y Y Y Y Y Y

Household ﬁxed eﬀects Y Y Y Y Y Y Y

Panel II: Contractionary episode

Rates rising recently 0.228*** 0.261*** 0.194*** 0.155*** 0.063*** -0.005 -0.019**

(0.016) (0.017) (0.016) (0.017) (0.016) (0.014) (0.009)

Controls Y Y Y Y Y Y Y

Quarter ﬁxed eﬀects Y Y Y Y Y Y Y

Household ﬁxed eﬀects Y Y Y Y Y Y Y

Notes: Results obtained from estimating equation (4) using data from the Canadian Survey of Consumer Expectations

(CSCE). ** and *** denote signiﬁcance levels at 5% and 1%. Standard errors are clustered at the consumer level.

Table 11: Responses of delinquencies (%) and credit scores

Mortgages Auto loans Installment loans Credit cards Lines of credit Credit

60-day 90-day 60-day 90-day 60-day 90-day 60-day 90-day 60-day 90-day score

Panel I: Expansionary episode

FRM-5yr

PostRenew -1.10*** -0.14 -0.13 0.00 -0.13 -0.23** 0.00 -0.44 0.02 -0.18 3.26***

(0.20) (0.07) (0.07) (0.00) (0.09) (0.11) (0.00) (0.26) (0.15) (0.12) (0.48)

FRM-4yr

PostRenew -0.13 -0.08 0.00 0.00 0.05 0.06 0.09 0.17 -0.12 -0.10 3.13***

(0.36) (0.17) (0.00) (0.00) (0.55) (0.16) (0.09) (0.30) (0.27) (0.20) (0.95)

FRM-3yr

PostRenew -0.23 0.07 0.00 0.00 -0.22 -0.05 -0.53 -0.06 0.05 0.13 1.62***

(0.19) (0.07) (0.00) (0.00) (0.13) (0.05) (0.34) (0.21) (0.17) (0.09) (0.59)

FRM-2yr

PostRenew 0.20 0.03 0.00 0.00 -0.02 -0.02 -0.13 0.02 0.02 -0.07 0.84**

(0.15) (0.07) (0.00) (0.00) (0.09) (0.08) (0.30) (0.17) (0.14) (0.08) (0.39)

Panel II: Contractionary episode

FRM-5yr

PostRenew -0.27 0.01 -0.06 0.00 -0.16 -0.09 -0.42 -0.17 -0.01 0.00 1.21**

(0.19) (0.06) (0.08) (0.07) (0.12) (0.06) (0.26) (0.15) (0.11) (0.07) (0.53)

FRM-4yr

PostRenew 0.24 0.08 0.00 -0.01 -0.20 0.00 0.40 0.16 0.12 -0.13 2.07***

(0.18) (0.07) (0.06) (0.03) (0.15) (0.09) (0.35) (0.21) (0.13) (0.09) (0.63)

FRM-3yr

PostRenew -0.40 0.05 -0.05 -0.09 0.22 -0.01 -1.10** -0.40 -0.32 0.02 0.19

(0.40) (0.15) (0.08) (0.08) (0.21) (0.10) (0.47) (0.32) (0.23) (0.16) (0.91)

FRM-2yr

PostRenew 0.12 0.01 -0.04 0.04 0.07 0.01 -0.44 -0.02 -0.18 -0.18** 0.36

(0.13) (0.06) (0.06) (0.03) (0.12) (0.07) (0.25) (0.15) (0.10) (0.08) (0.35)

Notes: Delinquencies are measured by the probability of reaching 60 days or 90 days of delinquency on at least one

account of a debt category. Each cell presents the results from estimating one regression using equation (1). ** and

*** denote signiﬁcance levels at 5% and 1%. Standard errors are clustered at the consumer level. All regressions

include a set of control variables (see Section 3), month ﬁxed eﬀects, and consumer ﬁxed eﬀects.

Table 12: Aggregate eﬀects of rate resets

Auto Durable Mortgage Revolving debt

(in billion dollars) spending consumption pay-down pay-down

Expansionary episode

All renewers 2.02 1.38 3.56 0

FRM-5yr renewers 1.55 1.06 2.71 0

Contractionary episode

All renewers 0 0 0.40 0.69

FRM-5yr renewers 0 0 0.36 0.30

Notes: Aggregate eﬀects obtained from applying formula (6). See detailed descriptions of the estimation procedure

for each category in Section 6.

Not-for-Publication Appendix

A. U.S. Jumbo Prime ARMs

Table A1: Summary statistics of jumbo prime 10-year interest-only ARMs originated in 2005-2007

Di Maggio et al. (2017) Our replication

BlackBox Logic data CoreLogic data

Borrower with Borrower with Borrower with Borrower with

ﬁve-year ARMs ten-year ARMs ﬁve-year ARMs ten-year ARMs

Mean SD Mean SD Mean SD Mean SD

FICO 723.3 39.4 736 39.7 737.8 42.5 743.8 41.6

Loan balance 357,949 271,600 536,342 347,622 568,253 415,143 652,719 427,126

LTV ratio 77.11 10.01 72.82 12.05 73.66 11.80 69.70 13.60

Initial interest rate 6.44 0.76 6.14 0.52 5.98 0.56 6.01 0.47

Fraction of loans originated in California 46.05 45.55

Fraction of loans originated in Florida 6.78 5.92

Fraction of loans originated in Virginia 5.91 4.57

Fraction of loans terminated within 5 years 50.12 43.26

Fraction of loans terminated within 5 years: Voluntary payoﬀ 35.82 38.01

Fraction of loans terminated within 5 years: Foreclosure 10.15 8.12

Number of borrowers 46,578 26,543 37,716 69,949

Source: CoreLogic; Di Maggio et al. (2017).

Di Maggio et al. (2017) provided a comprehensive analysis of the eﬀects of ARM resets on consumer

spending and mortgage repayment in the U.S. during the Great Recession. The sample underlying

their main results consists of mortgages in a speciﬁc segment of the U.S. mortgage market: jumbo

prime ARMs. Their study ﬁnds large, persistent increases in durable spending at the micro level,

which argues for the potency of the mortgage rate channel of monetary policy in stimulating the

U.S. economy. To assess the aggregate eﬀects of ARM resets and to provide policy prescriptions,

however, it is important to understand the representativeness of the mortgages in their study.

To shed light on this issue, we accessed the CoreLogic Private Label Securities-MBS dataset

through the Federal Reserve System’s RADAR Data Warehouse. The dataset covers over 90 percent

of the loans of prime jumbo securities in the market. We ﬁrst assess the share of the loans underlying

the main results of Di Maggio et al. (2017) in the U.S. mortgage market, i.e., jumbo prime 10-year

interest-only ARMs originated in 2005-2007 with an initial ﬁxed interest-rate period of 5 years. We

estimate that these loans accounted for about 1.8% of the overall U.S. mortgage originations at the

Another issue with assessing the aggregate eﬀects of ARM resets during periods of declining rates in the U.S. is

that long-term FRM borrowers would also take advantage of the low rates to reﬁnance their mortgages, confounding

the eﬀects of ARM resets on aggregate outcomes.

time.

We next examine loan characteristics of these ARMs. The summary statistics are shown in Table

A1. Overall, the distributions of the credit score, LTV ratio and initial interest rate are similar in

the two datasets. The main diﬀerence is the balances of 5-year ARMs, which are $570K on average

in our data, compared to $360K in the BlackBox Logic data. This discrepancy is not driven by the

outliers in our data. The 25th percentile of the balance distribution, for example, is $350K in our

data. Despite this diﬀerence, the average loan balances in both datasets are much larger than that

on representative agency FRMs, which was between $214K and $275K in 2005-2007, according to

Freddie Mac.

A somewhat surprising ﬁnding is the geographic concentration of this type of loans. California

alone accounted for 46% of its originations (compared to the state’s population share of 12%).

Another state that disproportionately originated these loans was Virginia. In addition, we ﬁnd

that these loans had high prepayment rates, about 50%. About 70% of the prepayments were

voluntary payoﬀs and 20% were foreclosures.

To conclude, the small share of these loans in the U.S. mortgage market and their distinct

features and geographic concentration suggest that the behavior of the corresponding borrowers may

not be representative of a typical American consumer who is likely to hold a 30-year FRM contract.

This also highlights that more research and broader evidence might be needed to understand the

rate reset channel of monetary policy.

This market share is computed as follows. The share of jumbo loans originated in 2005-2007 is close to 30% of

the market (see, FDIC Quarterly Volume 13(4), 2019). In our prime-jumbo sample, 10-year interest-only ARMs over

the same period accounted for 20%, of which loans with an initial ﬁxed interest-rate period of 5 years are about 30%.

The market share, therefore, is 1.8% (=30%×20%×30%).

B. Construction of Mortgage Rates

We take a series of steps to impute the rates associated with the FRMs in our sample. First,

assuming no prepayment in addition to scheduled payments, the outstanding balances and scheduled

payments can be used to pin down the mortgage rate (adjusted to annual rate). Second, from the

rates obtained in the ﬁrst step, we remove the ones that are either too low (most likely due to

prepayment beyond the amortization schedule) or too high (most likely due to delays in payments).

Third, we take the median of the remaining rates within each term of a mortgage as the contracted

rate. Finally, we winsorize our contracted rates using the 1% cutoﬀ at the top and bottom the

distribution. A minor caveat of this procedure is that we are unable to recover the rates for a small

fraction of loans that feature either systematic prepayment or frequent delays in payments.

To externally validate our procedure, we compare the imputed rates in our data with the actual

rates reported in two datasets. One is the average 5-year FRM rate quoted by national mortgage

brokers, and the other is the loan-level rate recorded in the Bank of Canada (BOC)-OSFI mortgage

originations dataset. The broker data span a long time period, but do not have a cross-sectional

dimension. The BOC-OSFI dataset allows us to further break down the data by mortgage insurance

status and purpose, but is available only since 2014. Mortgages in both sources are new originations,

so we compare their rate distributions with those of the new originations in our sample.

Figure B1 shows that the average and median of our imputed rates track the brokers’ rate

quite closely over time. Classifying mortgages by their insurance status, Figure B2 shows that the

imputed rates are similar to the rates in the BOC-OSFI dataset, and that the rate diﬀerentials

for insured and uninsured mortgages are small. Although our sample does not distinguish between

loan purposes, the BOC-OSFI dataset suggests that the rates for home purchases and for cash-out

reﬁnances are similar, especially for uninsured mortgages. We also compare the standard deviations

of our imputed rates with those in the BOC-OSFI dataset by insurance status. The standard

deviations are similar, with a gap of about 20-30 bps over the sample period.

Figure B1: Imputed TransUnion rates and national mortgage brokers’ rate, 5-year FRMs

2 3 4 5 6

percent

2008m1 2010m1 2012m1 2014m1 2016m1 2018m1

Average TU rate Median TU rate Average brokers’ rate

Sources: Bank of Canada mortgage broker rate; TransUnion Canada account-level data; authors’ calculations.

Figure B2: Imputed TransUnion rates and BOC-OSFI origination rates, 5-year FRMs

2 2.5 3 3.5 4

percent

2014m1 2015m1 2016m1 2017m1 2018m1

Median TU rate Median OSFI rate (all purposes)

Median OSFI rate (purchases)

Uninsured mortgages

2 2.5 3 3.5 4

percent

2014m1 2015m1 2016m1 2017m1 2018m1

Median TU rate Median OSFI rate (all purposes)

Median OSFI rate (purchases)

Insured mortgages

Sources: Bank of Canada-OSFI mortgage originations dataset; TransUnion Canada account-level data; authors’

calculations.

C: Canadian Survey of Consumer Expectations

The Canadian Survey of Consumer Expectations (CSCE), launched by the Bank of Canada in

2014Q4, provides comprehensive information about consumer expectations of inﬂation, interest

rates, labor markets, credit markets and housing markets. The survey also collects information on

households’ demographics, employment situation and ﬁnancial conditions. Data are collected from

a nationally representative sample of 1,000-2,000 household heads every quarter. The survey has a

rotating panel structure with respondents participating in the panel for up to a year. The design

of the survey largely follows that of the New York Fed’s Survey of Consumer Expectations. See

Gosselin and Khan (2015) for a detailed description of the CSCE.

For our analysis in Section 5.4, we restrict the sample to homeowners with mortgages to make

it comparable with our consumer credit panel data. Summary statistics of the regression variables

are presented in Table C1. The following survey questions are used to construct the key variables

in the regressions.

1. “How would you say interest rates on things such as mortgages, bank loans and savings have

changed over the last 12 months? Fallen a lot, fallen a little, about the same, risen a little,

risen a lot?” We use the answer to this question to construct an indicator variable for

perceiving that interest rates have risen. The indicator equals one if the consumer answers

risen a lot or risen a little.

2. “What do you think is the percent chance that 12 months from now the average interest rate

on things such as mortgages, bank loans and savings will be higher than it is now?” We use

the answer to this question to construct an indicator variable for expecting the rates to be

higher in the next 12 months. The indicator equals one if the consumer gives a numerical

answer greater than 50%.

3. “At what level do you think that interest rates on things such as mortgages, bank loans and

savings will be in...?” Please enter a number.

(a) One year from now, interest rates will most likely be __%

(b) Two years from now, interest rates will most likely be __%

We construct an indicator for a consumer expecting rates to rise in the next two years, which

equals one if the consumer (i) expects rates to be higher in the next 12 months (based on the

answer to question 2), and (ii) reports the level of the expected rate in two years (answer to

question 3b) greater than the level of the expected rate in one year (answer to question 3a).

An indicator for expecting rates to rise in the next ﬁve years is constructed similarly, equal

to one if the consumer meets conditions (i) and (ii), and reports the level of the expected rate

in ﬁve years (question 3c) greater than the level of the expected rate in two years (question

3b).

4. “Which, if any, of the following actions are you taking, or planning to take, in light of your

expectations for interest rates?” Select all that apply.

• Bring forward major purchases (such as furniture or appliances)

• Postpone major purchases

• Cut back spending and save more

• Pay debt

• Take no action

We construct a series of indicators for taking a speciﬁc action.

Table C1: Summary statistics of CSCE variables

Both episodes Expansionary Contractionary

Mean SD Mean SD Mean SD

Household characteristics

Age 44.8 13.1 44.4 12.9 45.1 13.2

Male (%) 50.0 50.0 50.0 50.0 50.0 50.0

Married (%) 74.3 43.7 74.9 43.3 73.9 43.9

Bachelor degree or above (%) 47.9 49.9 47.9 49.9 47.9 49.9

Existing FRM borrowers (%) 69.4 46.1 67.3 46.9 70.8 45.4

Existing VRM borrowers (%) 21.9 41.3 23.3 42.3 20.9 40.7

New FRM/VRM borrowers (%) 8.7 28.2 9.4 29.2 8.2 27.5

Expectations

Perceiving rates risen in past 12M (%) 53.8 49.9 18.1 38.5 75.8 42.8

Expecting rates higher in next 12M (%) 67.0 47.0 56.3 49.6 73.7 44.0

cond. on perceiving rates risen (%) 79.7 40.2 74.2 43.8 80.6 39.6

cond. on perceiving rates fallen (%) 51.0 50.0 50.9 50.0 51.1 50

Expected interest rates in one year (pp) 5.5 7.5 4.4 4.9 6.3 9.1

Expected interest rates in two years (pp) 6.6 8.6 5.2 5.5 7.3 9.8

Expected interest rates in ﬁve years (pp) 7.4 9.0 6.5 7.3 8.2 10.4

Actions

Pay down debt (%) 53.1 49.9 52.9 49.9 53.2 49.9

Cut spending/save more (%) 44.3 49.7 38.5 48.7 47.9 49.9

Postpone purchases (%) 18.5 38.9 15.3 36.0 20.5 40.4

Bring forward purchases (%) 7.3 25.9 6.3 24.2 7.9 26.9

Notes: Summary statistics for homeowners with mortgages in the CSCE data. Statistics on the expected interest

rates in one, two and ﬁve years are computed after dropping the top 1% of the sample in order to reduce the eﬀect

of outliers.

D. Additional Empirical Evidence and Robustness

Table D1: Term transition probabilities and market shares (%)

After reset

FRM-2yr FRM-3yr FRM-4yr FRM-5yr Share

Before reset

Panel I: Expansionary episode

FRM-2yr 65.3 6.8 4.2 23.6 30.0

FRM-3yr 36.7 20.1 5.7 37.5 9.3

FRM-4yr 22.5 6.3 18.3 52.9 6.2

FRM-5yr 19.0 7.5 5.1 68.4 54.5

Panel II: Contractionary episode

FRM-2yr 57.5 16.0 6.8 19.7 28.3

FRM-3yr 29.8 34.5 7.7 28.1 14.8

FRM-4yr 24.1 13.4 27.7 34.8 13.4

FRM-5yr 20.1 12.4 11.3 56.2 43.5

Source: TransUnion Canadian mortgage account-level data; authors’ calculations.

Table D2: Borrowing activity following loan termination

Expansionary episode Contractionary episode

Share with with Share with with

current lender other lender current lender other lender

Loans terminated before scheduled renewal 28% - - 26% - -

Among terminated loans:

1. Replaced by a cash-out-reﬁnance loan 46% 86% 14% 40% 78% 22%

(Balance rises by 5% or more; same postal code)

2. Replaced by a renewal loan 10% 38% 62% 13% 22% 78%

(Balance rises by 0-5%; same postal code)

3. Replaced by a home-purchase loan 5% 66% 34% 5% 57% 43%

(New account opened in a diﬀerent postal code)

4. No replacement loan found 39% - - 43% - -

Notes: These estimates are obtained by linking the mortgage accounts of the same borrowers. Speciﬁcally, for each

terminated loan, we search for a replacement loan originated within three months of the termination date of the

original loan by the same borrower. The replacement loan may be issued by any lender in the TransUnion database.

This table presents the estimates using all mortgages. The estimates by loan term are similar to those shown in this

table.

Table D3: Changes in monthly debt service payments (in dollars) upon the reset

Total Mortgage Non-mortgage Revolving debt

(1) (2) (3) (4)

Panel I: Expansionary episode

FRM-5yr

PostRenew -60.22*** -46.47** -14.7*** -6.76***

(2.14) (0.64) (2.00) (0.94)

FRM-4yr

PostRenew -9.42*** -9.90*** 0.33 -3.52

(4.64) (1.85) (4.31) (2.04)

FRM-3yr

PostRenew -11.15*** -2.19 -8.91*** -6.05***

(2.68) (1.17) (2.43) (1.23)

FRM-2yr

PostRenew -8.41*** -1.76** -6.76*** -4.49***

(1.74) (0.88) (1.51) (0.77)

Panel II: Contractionary episode

FRM-5yr

PostRenew 22.92*** 39.23*** -15.91*** -8.36***

(2.52) (0.73) (2.41) (1.34)

FRM-4yr

PostRenew 32.51*** 40.37*** -7.75*** -4.83***

(2.88) (0.77) (2.77) (1.53)

FRM-3yr

PostRenew 33.29*** 49.49*** -15.93*** -7.93***

(3.80) (1.31) (3.58) (2.04)

FRM-2yr

PostRenew 75.53*** 84.49*** -7.62*** -5.86***

(1.54) (0.85) (1.29) (0.77)

Notes: Each cell presents the results from estimating one regression using equation (1). ** and *** denote signiﬁcance

levels at 5% and 1%. Standard errors are clustered at the loan level. All regressions include a set of control variables

(see Section 3), month ﬁxed eﬀects, and loan ﬁxed eﬀects. Column (1) shows the responses of total monthly debt

service payments, which include scheduled mortgage payments (column 2) and required non-mortgage payments

(column 3). The latter consist of required payments on revolving debt (column 4) and on non-revolving debt (auto

and installment loans).

Table D4: Robustness: Post-renewal observations restricted to be within four quarters

Mortgage Required Scheduled Amortization Auto spending Auto pur. IL-Financed IL pur. Revolving CC LOC Mortgage

rate (p.p.) ($/m) ($/m) (months) ($/m) prob. (%) spending ($/m) prob. (%) ($) ($) ($) 60-day (%)

Panel I: Expansionary episode

FRM-5yr

Renew× -0.96*** -80.53*** -48.25*** -8.96*** 27.63** 0.08*** 40.59 0.18*** 319.36 -146.36*** 494.85** -0.53

PostRenew (0.008) (1.06) (1.52) (0.35) (11.44) (0.03) (22.88) (0.05) (217.83) (51.89) (211.04) (0.42)

FRM-4yr

Renew× -0.19*** -21.26*** -5.85 -4.30*** -71.76** -0.18 30.06 0.17 959.95 -184.71 1083.12 -0.18

PostRenew (0.014) (1.63) (3.14) (0.69) (33.64) (0.09) (58.57) (0.15) (577.57) (107.17) (579.40) (0.71)

FRM-3yr

Renew× -0.12*** -8.60*** 0.34 -3.78*** 16.46 0.03 36.07 0.04 376.14 -151.56*** 590.65 0.31

PostRenew (0.001) (1.10) (2.03) (0.35) (18.44) (0.05) (30.29) (0.07) (304.81) (58.87) (302.26) (0.42)

FRM-2yr

Renew× -0.04*** -5.93*** 0.99 -3.88*** -5.09 -0.03 36.21 0.01 349.85 -172.03*** 595.67*** 0.67**

PostRenew (0.008) (0.85) (1.76) (0.29) (17.34) (0.05) (33.30) (0.07) (197.42) (42.09) (195.27) (0.33)

Panel II: Contractionary episode

FRM-5yr

Renew× 0.31*** 32.68*** 38.66*** -1.84*** 0.38 0.02 49.57** 0.07 -1100.00*** -289.32*** -937.53*** -0.23

PostRenew (0.005) (0.60) (0.98) (0.12) (10.19) (0.03) (21.94) (0.05) (255.49) (50.58) (249.80) (0.24)

FRM-4yr

Renew× 0.70*** 52.60*** 58.32*** -2.30*** -16.92 -0.05 -8.55 0.02 -784.39 -291.53*** -510.80 0.33

PostRenew (0.007) (0.79) (1.40) (0.22) (19.75) (0.05) (42.05) (0.09) (415.78) (83.75) (400.41) (0.36)

FRM-3yr

Renew× 0.92*** 72.81*** 63.95*** 1.07*** 11.85 0.03 52.47 0.13 -612.37 -130.91 -440.56 0.09

PostRenew (0.011) (1.52) (2.13) (0.36) (27.53) (0.07) (45.49) (0.12) (490.10) (94.27) (470.46) (0.58)

FRM-2yr

Renew× 1.02*** 101.78*** 102.55*** -2.16*** 9.82 0.03 78.13** 0.14** -8.87 -210.56*** 181.24 0.28

PostRenew (0.004) (1.14) (1.46) (0.16) (15.78) (0.04) (32.39) (0.06) (206.08) (36.57) (201.70) (0.21)

Notes: See notes in Tables 3 and 6 for estimation details. These estimates are obtained using the sample that restricts the post-renewal observations to be within

four quarters for all loans.

Table D5: Robustness: Mortgage terms restricted to be the same before and after the reset

Mortgage Required Scheduled Amortization Auto spending Auto pur. IL-Financed IL pur. Revolving CC LOC Mortgage

rate (p.p.) ($/m) ($/m) (months) ($/m) prob. (%) spending ($/m) prob. (%) ($) ($) ($) 60-day (%)

Panel I: Expansionary episode

FRM-5yr

PostRenew -1.14*** -96.77*** -43.39*** -15.79*** 13.89 0.07*** 47.20*** 0.12*** 168.42 -188.58*** 363.54** -0.10

(0.004) (0.61) (1.00) (0.25) (7.92) (0.02) (14.82) (0.04) (156.66) (41.46) (152.02) (0.15)

FRM-4yr

PostRenew -0.36*** -32.88*** -7.53 -5.41*** -55.75 -0.03 40.58 0.22 -2.16 -77.58 -55.94 -0.08

(0.012) (1.75) (5.01) (0.89) (58.83) (0.14) (74.33) (0.20) (667.80) (166.23) (668.86) (0.21)

FRM-3yr

PostRenew -0.21*** -19.45*** 3.68 -5.78*** 26.14 0.07 39.15 0.14 -583.03 -255.18*** -195.81 0.01

(0.007) (0.77) (2.46) (0.50) (25.61) (0.07) (42.54) (0.11) (476.96) (92.17) (469.25) (0.30)

FRM-2yr

PostRenew -0.35*** -31.92*** -16.44** -4.72*** 2.69 0.00 12.65 0.06 149.94 -162.33*** 348.74** -0.02

(0.003) (0.33) (1.12) (0.23) (13.81) (0.04) (25.31) (0.06) (165.54) (33.05) (163.30) (0.09)

Panel II: Contractionary episode

FRM-5yr

PostRenew 0.30*** 30.01*** 38.91*** -1.77*** 10.62 0.02 12.50 0.04 -436.33 -218.36*** -319.07 -0.04

(0.004) (0.51) (1.01) (0.15) (12.39) (0.03) (21.69) (0.05) (289.25) (57.33) (286.15) (0.08)

FRM-4yr

PostRenew 0.47*** 32.66*** 39.34*** -0.99*** 33.67 0.09 -6.15 0.14 -222.16 -338.29*** 213.29 0.11

(0.004) (0.50) (1.19) (0.22) (22.81) (0.07) (61.22) (0.11) (439.61) (104.23) (430.48) (0.09)

FRM-3yr

PostRenew 0.64*** 51.81*** 48.00*** 1.11** -24.91 -0.08 75.56 0.01 -632.62 -325.58*** -290.63 -0.11

(0.009) (1.39) (2.48) (0.43) (31.28) (0.09) (54.46) (0.13) (583.98) (121.31) (565.55) (0.11)

FRM-2yr

PostRenew 0.68*** 66.37*** 65.43*** -0.07 4.10 0.05 23.78 0.07 -142.25 -171.09*** 77.62 -0.00

(0.003) (0.81) (1.15) (0.17) (15.89) (0.04) (35.93) (0.06) (195.73) (35.16) (194.98) (0.09)

Notes: See notes in Tables 3 and 6 for estimation details. These estimates are obtained using the sample that restricts the mortgage term to be the same before

and after the reset.

Table D6: Robustness: DID estimates (Control group: longer-term FRMs not yet renewed)

Mortgage Required Scheduled Amortization Auto spending Auto pur. IL-Financed IL pur. Revolving CC LOC Mortgage

rate (p.p.) ($/m) ($/m) (months) ($/m) prob. (%) spending ($/m) prob. (%) ($) ($) ($) 60-day (%)

Panel I: Expansionary episode

FRM-5yr

Renew× -1.13*** -92.04*** -46.50*** -13.95*** 19.46*** 0.07*** 45.54*** 0.14*** 101.74 -162.31*** 254.23** -1.09***

PostRenew (0.004) (0.54) (0.81) (0.20) (6.09) (0.02) (11.98) (0.03) (124.83) (32.46) (121.02) (0.20)

FRM-4yr

Renew× -0.36*** -32.28*** -9.14*** -5.75*** -6.06 -0.00 3.92 0.07 -1000*** -314.57*** -708.86 -0.15

PostRenew (0.006) (0.76) (1.72) (0.34) (16.34) (0.04) (26.51) (0.07) (373.70) (72.04) (370.44) (0.30)

FRM-3yr

Renew× -0.19*** -15.05*** -3.40*** -4.35*** 9.49 0.03 18.48 0.05 -964.16*** -287.54*** -630.19*** -0.11

PostRenew (0.004) (0.50) (1.13) (0.21) (8.38) (0.02) (14.57) (0.03) (195.76) (38.88) (192.36) (0.16)

FRM-2yr

Renew× -0.18*** -15.23*** -2.55** -4.88*** 4.55 0.02 25.79 0.08*** -384.78** -154.17*** -206.23 0.33**

PostRenew (0.003) (0.38) (0.87) (0.18) (7.99) (0.02) (15.89) (0.03) (159.76) (31.87) (156.69) (0.13)

Panel II: Contractionary episode

FRM-5yr

Renew× 0.31*** 33.87*** 38.88*** -1.54*** 3.08 0.02 15.19 -0.01 -795.41*** -214.97*** -666.28*** -0.19

PostRenew (0.003) (0.44) (0.72) (0.11) (7.65) (0.02) (14.46) (0.03) (217.86) (41.18) (212.79) (0.17)

FRM-4yr

Renew× 0.49*** 36.32*** 40.35*** -1.07*** -3.23 -0.03 -0.40 0.06 86.23 -141.74*** 232.61 0.43

PostRenew (0.003) (0.34) (0.77) (0.13) (12.97) (0.03) (23.63) (0.06) (269.87) (53.34) (261.11) (0.45)

FRM-3yr

Renew× 0.70*** 55.32*** 49.68*** 0.72*** 16.64 0.04 48.70 0.12 -826.07** -272.99*** -498.50 0.93

PostRenew (0.006) (0.77) (1.31) (0.24) (18.57) (0.05) (27.72) (0.07) (329.19) (65.25) (318.38) (0.72)

FRM-2yr

Renew× 0.85*** 83.31*** 84.43*** -1.36*** 17.68 0.06** 41.53 0.07 -304.88** -216.37*** -75.70 -0.03

PostRenew (0.003) (0.66) (0.85) (0.12) (10.50) (0.03) (23.29) (0.04) (136.37) (24.91) (134.54) (1.00)

Notes: Each cell presents the results from estimating one regression using equation (3). ** and *** denote signiﬁcance levels at 5% and 1%. Standard errors are

clustered at the loan (consumer) level. All regressions include a set of control variables (see Section 3), month ﬁxed eﬀects, and loan (consumer) ﬁxed eﬀects. The

control group consists of 7- and 10-year FRMs that previously reset the rates at the same time as the treatment group but were not scheduled to be renewed in

the episode.

Table D7: Robustness: DID estimates (Control group: FRMs having the same term as the treated group but not yet renewed)

Mortgage Required Scheduled Amortization Auto spending Auto pur. IL-Financed IL pur. Revolving CC LOC Mortgage

rate (p.p.) ($/m) ($/m) (months) ($/m) prob. (%) spending ($/m) prob. (%) ($) ($) ($) 60-day (%)

Panel I: Expansionary episode

FRM-5yr

Renew× -1.15*** -93.87*** -49.62*** -13.39*** 11.00** 0.04*** 19.20** 0.08*** -1600*** -384.37*** -1200*** -1.17***

PostRenew (0.003) (0.51) (0.76) (0.18) (4.33) (0.01) (8.28) (0.02) (110.54) (26.04) (108.03) (0.12)

FRM-4yr

Renew× -0.34*** -29.61*** -7.88*** -5.54*** 7.37 0.03 29.58 0.11** -2000*** -416.09*** -1600*** -0.17

PostRenew (0.006) (0.71) (1.61) (0.31) (11.13) (0.03) (17.99) (0.05) (270.02) (53.06) (266.82) (0.23)

FRM-3yr

Renew× -0.19*** -15.17*** -3.38*** -4.43*** 12.82 0.03 20.63 0.03 -557.48*** -202.41 -288.73*** -0.27

PostRenew (0.004) (0.49) (1.12) (0.21) (9.18) (0.03) (15.85) (0.04) (177.70) (35.32) (174.11) (0.16)

Panel II: Contractionary episode

FRM-5yr

Renew× 0.31*** 33.09*** 37.63*** -1.53*** 6.42 0.01 20.83 0.04 -2000*** -451.93*** -1600*** -0.54***

PostRenew (0.003) (0.41) (0.67) (0.10) (5.53) (0.01) (12.89) (0.03) (163.02) (31.41) (160.10) (0.13)

FRM-4yr

Renew× 0.52*** 38.70*** 42.13*** -1.11*** 8.11 0.00 7.25 0.07 19.10 -278.76*** 305.32 -0.03

PostRenew (0.003) (0.34) (0.71) (0.12) (12.29) (0.03) (23.14) (0.05) (223.93) (47.87) (217.88) (0.14)

FRM-3yr

Renew× 0.73*** 58.87*** 52.42*** 0.83*** -1.10 0.00 57.72** 0.13** -1500*** -330.88*** -1100*** -0.23

PostRenew (0.005) (0.79) (1.29) (0.23) (14.06) (0.04) (25.06) (0.06) (262.03) (52.24) (254.71) (0.31)

Notes: Each cell presents the results from estimating one regression using equation (3). ** and *** denote signiﬁcance levels at 5% and 1%. Standard errors are

clustered at the loan (consumer) level. All regressions include a set of control variables (see Section 3), month ﬁxed eﬀects, and loan (consumer) ﬁxed eﬀects. The

control group consists of mortgages that had the same terms as the treatment group but were not scheduled to be renewed in the episode.