The Psychological Cost of Debt:

Evidence from Housing Mortgages in Pakistan*

Saad Azmat„

Isabel Harbaugh Macdonald…(Job Market Paper)

September 15, 2020

Abstract

Holding debt imposes both monetary and mental costs on the debtor. How do bor-
rowers prioritize financial versus psychological costs, and what do these choices tell us
about the mental burden of debt? In this project, we explore these trade-offs using
repayment data for 3,705 borrowers with an Islamic housing mortgage in Pakistan.
The product allows borrowers to make early payments that reduce anxiety but often
increase the cost of borrowing. Forty percent of borrowers in our sample make early
payments, leading to an average cost of PKR 354,155 ($4,132 USD) in additional fees
and foregone savings. We develop a new model of consumer hedonics to explain these
results, and we rule out religion, demand for commitment devices, and misunderstand-
ing of costs as complete explanations. These findings suggest that the mental burden
of debt is high, and that there is a psychological benefit to decreasing debt, even if an
account is not closed fully.

* We thank John Beshears, Will Dobbie, Asim Khwaja, Blake Heller, Augustin Bergeron, and Kunal Man-
gal as well as seminar participants at Harvard University and Lahore University School of Management for
helpful comments. This material is based upon work supported by the National Science Foundation Grad-
uate Research Fellowship Program. Any opinions, findings, and conclusions or recommendations expressed
in this material are those of the authors and do not necessarily reflect the views of the National Science
Foundation. This study was approved by the Harvard University Institutional Review Board (IRB18-1645).
„ Lahore University of Management Sciences (email: [email protected])
… Harvard University (email: [email protected])
1 Introduction

Holding debt is costly. American consumers spend about $500 billion on interest payments
for housing mortgages alone.1 In lower income countries, reliance on mortgages and other
debt contracts is expanding along with the middle class, and interest rates are often above
10% annually. Financial costs, however, are not the only burden of debt. Higher debt is
associated with mental costs like anxiety, psychological distress, and depression (Gathergood
2012, Sweet et al. 2013, Richardson et al. 2013). Attending to debt also imposes a “bandwidth
tax” on mental resources, which has been shown to reduce cognitive functioning (Mani et al.
2013, Ong et al. 2019). This evidence raises the question of whether psychological costs may
sometimes exceed the financial burden of indebtedness.
The twin costs of debt—financial and psychological—often act in tandem to push con-
sumers towards reducing their debt burden, making it difficult to separate their effects. In
some contexts, however, financial and psychological costs run in opposite directions. When
faced with the option to reduce one cost and increase the other, how do borrowers decide?
What can these choices tell us about the unobservable mental cost of debt? In this project,
we directly compare financial and psychological costs of debt in a novel field context. We an-
alyze repayment behavior for 3,705 consumers in Pakistan with an Islamic housing mortgage
called a diminishing musharakah. The product functions similarly to a standard mortgage
loan but it contains a special feature that allows borrowers to add “early payments” with
each installment that eliminate their debt more quickly. Faster repayment may reduce the
anxiety of holding debt, but these early payments also bring additional costs. Borrowers
are charged a 3% fee on repaid principal and forego earnings from keeping funds in a high-
interest savings account, which often outweighs any savings from reduced interest on the
loan.
The diminishing musharakah loan is an ideal field setting to study the dual costs of debt
because with early payments, borrowers trade monetary losses for psychological gains. With
1
Estimation based on data for 2019 from the Federal Reserve and Fannie Mae.

1
other debt products, psychological and financial costs increase or decrease simultaneously,
which makes it difficult to separately identify their role in shaping preferences and behavior.
In our study, decreasing stress with early payments increases financial costs, allowing us to
quantify the mental costs of debt through the amount borrowers are willing to pay for psy-
chological relief. Furthermore, the early payment feature provides a low-friction mechanism
for borrowers to express their desire to repay debt early. Many mortgage products allow
early payments, but due to high transaction costs, most borrowers only repay early if they
are refinancing an entire loan. With the diminishing musharakah loan, borrowers can repay
as little as one extra month and have no added paperwork, leading to a dynamic measure of
desire for early repayment.
In the first part of the paper, we establish four facts about repayment behavior using
data from the diminishing musharakah sample. In Fact 1, we show that early payments are
frequent and sizable. Forty percent of borrowers make at least one early payment over the
course of the fifteen year sample period. Since some loans are ongoing, it is likely that a
higher proportion of borrowers will make early payments before their loans are complete. On
average, early payers make 3.3 early payments totaling 85 months of principal repaid early, so
the average early payment includes 25.7 months of principal (called “units”). Early payments
can be as small as one unit repaid early, so the variance on payment size is substantial.
In Fact 2, we establish that early payments can be financial costly to consumers. Bor-
rowers who make at least one early payment in the sample lose an average of PKR 354,155
($4,132 USD) in fees and foregone savings account earnings over the course of their loans.
If we convert these costs to months of principal using the average loan size in our sample,
early payers lose the equivalent of 17.6 months of principal by making early payments in-
stead of keeping money in savings and paying on the regular schedule. The costs of each
payment vary based on prevailing interest rates, but they are highest in the first quarter
of loan tenure and lowest in the third quarter. One extra month (one unit) repaid in the
first quarter of loan tenure costs early payers an average of PKR 5,967 ($70 USD), while an

2
equivalent repayment in the third quarter cost PKR $1,070 ($12 USD) on average. In Fact
1, we showed that early payers repay 85 units early on average, which leads to the average
total cost of PKR 354,155 ($4,132 USD).
In Fact 3, we show that early payments are more common early in the loan tenure. Among
subjects who have fully repaid their loans in our sample period, 66% of early payments are
made in the first half of loan tenure (e.g. the first ten years of a 20-year mortgage). This
finding is surprising because in Fact 2 we show costs are higher earlier in the loan tenure.
While the average cost per unit was nearly six times higher for payments in the first quarter
versus the third quarter, almost 60% more early payments were made in the first quarter
relative to the third quarter.
In Fact 4, we show that few early payments are made to fully close out a loan. We
analyze repayment behavior in the final three years of actual installments and see no spike
in payments at the final installment or in the final year. Only 3% of early payments are
made within the last three months that a loan account is open, relative to 6% of payments
being made in the first three months of a loan. This finding contrasts with prior theories
that debt averse behavior is driven by a desire to fully close out accounts, which is explored
more closely in subsequent sections (e.g. Amar et al. 2011).
To organize these facts, we develop a model of consumer hedonics around debt repayment.
We show that a standard utility model with no debt aversion cannot explain Facts 1 and 2
that early payments are frequent and costly. We next test a model of debt account aversion
in which holding debt causes stress and subjects can increase utility by fully closing a debt
account. This model aligns with prior lab findings that people often willing to pay more in
borrowing costs to close out small balance accounts (Amar et al. 2011, Kettle et al. 2016).
Debt account aversion is consistent with the existence of costly early payments in Facts 1 and
2 but cannot explain Facts 3 and 4 that most early payments are made early in a loan. We
adjust the model to add debt level aversion in which people gain utility by proactively paying
down debt, regardless of whether the account is closed. This model appears consistent with

3
all the repayment facts presented.
Using this final model, we identify additional predictions about debt preferences to in-
vestigate in our data in order to provide extra support for the model. We supplement these
tests with results from a survey of 63 diminishing musharakah borrowers from our sample
on their perceptions and demand for early payments. We show that as the model predicts,
early payers in the survey report high levels of debt-related stress and that early payments
bring them positive emotions of relief and pride. Borrowers who self-report making early
payments are 24 percentage points more likely to say debt makes them uncomfortable and
20 percentage points more likely to say debt causes them stress relative to non-early payers.
Among early payers, 80% say early payments make them feel relieved and 80% say early
payments make them feel proud, suggesting positive affective impacts from making early
payments.
We also confirm that debt level aversion and debt account aversion are separable effects,
as the model predicts. We collect a measure of debt account aversion from our survey sample
using a one-period version of the lab game from Amar et al. (2011), which we compare to
self-reported early payments. We find that respondents who made early payments are 30
percentage points less likely to prioritize closing the lowest balance accounts, which is the
repayment strategy most closely associated with debt account aversion. Early payers are
17 percentage points more likely to prioritize the highest interest account, which technically
would allow them to repay debt fastest because they would have more money to allocate to
principal payments rather than interest. These findings suggest that debt account aversion
in earlier studies may be different than the debt level aversion we observe in our context
through early payments, which reflects a desire to reduce the integral on outstanding debt,
even when full repayment is years in the future.
Last, the model predicts that early payments will be less frequent in periods when people
face lower earnings and liquidity, which is supported by our data. Borrowers with install-
ments due during the last week of the month are 11 percentage points less likely to make

4
any early payments, and 25 percentage points less likely to make small early repayments of
6 extra months or less. Most borrowers are paid monthly and likely face higher liquidity
constraints at the end of the month, so this finding suggests that liquidity concerns lower the
rate of early payments. Borrowers also make more payments in early months of the years
when salary increases take effect and they are likely to have more cash on hand.
In the final part of the paper, we test several alternate explanations for our repayment
facts and find that they cannot fully explain our facts. We show religious pressure is unlikely
to drive payments because early payments are less frequent during times of religious salience
and early payers in the survey are no more likely to say that debt should be reduced for
religious reasons. Similarly, we conclude early payments are not explained by demand for
commitment devices because borrowers have more profitable alternatives like term deposit
savings accounts that offer higher interest in exchange for multi-year savings commitment.
Early payers and non-early payers are also equally likely to say that having excess cash in
savings would cause them to spend unnecessarily or give money to friends/family, showing
consistent demand for commitment vehicles across groups.
We also find evidence that misunderstanding of costs do not fully explain early payments,
though we cannot determine if people have accurately calculated the full cost of these pay-
ments. Borrowers appear to be generally aware that early payments entail financial costs
because 87% say that early payments are “responsible” even if there is a financial cost to
making them. Early payers also express a willingness to accept up to a 6% principal fee
on average, which is double the actual fee charged on early payments. We therefore con-
clude that people are aware of the directionality of financial costs, but we cannot determine
definitively that they have accurately calculated the magnitude.
This project contributes to our understanding of the trade-offs between financial and
psychological costs of debt in several ways. First, the context allows us to quantify the
psychological cost of debt via revealed preference for early debt payments over savings.
Other studies have documented that debt appears associated with higher levels of stress,

5
both in terms of correlation (see Richardson et al. 2013 for a meta-review) and causality
(Gathergood 2012, Ong et al. 2019). Without a way to quantify the impact of debt on stress
in dollar terms, however, it is difficult to assess how much the rational debtor should pay
money to reduce psychological costs, or whether people should prioritize cost minimization
above all else in deciding how to repay debt. In our study, early payers chose to forgo over
$4,000 USD to repay early, and lower debt-related anxiety appears to be the primary benefit.

While the magnitude of this estimate may be impacted by miscalculations in the costs of
early payments, the general awareness of financial costs and stated willingness to pay higher
fees that we found in our survey show that early payers are consciously choosing to reduce
stress through early payments in exchange for a higher cost of borrowing.
This finding highlights the importance of mental costs in the growing literature on optimal
debt repayment. Several studies have documented that borrowers often fail to minimize costs
when repaying debts (Gathergood et al. 2019, Keys et al. 2016, Ponce et al. 2017, Stango
and Zinman 2009). Our study shows that the mental burden of holding debt and mental
benefits from repayment can be a contributing factor to repayment strategies that appear
sub-optimal. To help consumers optimize their debt repayment and increase overall well-
being, it will is critical to account for the weighty psychological impacts of debt and help
borrowers adopt strategies that decrease both the mental and financial costs of debt.
Second, our findings expand the conception of debt aversion to include a preference to
pay down debt even when an account remains open. Prior studies have shown debt aversion
in people’s preference to close out small debt accounts rather than minimize borrowing costs
(Amar et al. 2011, Besharat et al. 2014, Kettle et al. 2016). Studies typically attribute
this behavior to motivational impacts and the desire to achieve a tangible sense of progress
towards full debt repayment, although the reduction in transaction costs from reducing open
accounts also seems important. We find little evidence of debt account aversion in our
setting because few early payments are made to fully close a loan. This divergence with
prior literature may arise because borrowers are charged a 3% fee on early payments in our

6
setting, which may suppress the demand to repay faster when full repayment is proximate.
Furthermore, earlier work on debt account aversion analyzes repayment across multiple small
loans, while we only observe repayment of one loan that is likely the largest in a consumer’s
portfolio.
Third, studies on debt repayment have focused largely on credit card debt and other
debt accounts that are within a few years of repayment (Gathergood et al. 2019, Ponce et al.
2017). Our study explores long-term debt repayment and captures trends that may not be
apparent in short-term liabilities. We find that early payments are most common at the
start of the loan, suggesting that the psychological cost of debt might be highest in the early
stages of repayment. Our findings are especially pertinent to the study of housing-related
debt because loans have a long tenure and psychological forces like the fear of foreclosure or
the pride in home ownership are strong and may influence attitudes towards repayment.
The remainder of the paper proceeds as follows. In Section 2, we provide additional
details on the loan context, data, and sample. In Section 3, we establish facts from our
data on the costs and prevalence of early repayments. In Section 4, we present a model of
consumer hedonics to explain these facts. In Section 5, we provide supporting evidence for
implications from the model. In Section 6, we rule out alternative explanations for our facts.
Section 7 concludes.

2 Data and Context

This section first discusses our primary data sample of diminishing musharakah repayments.
We provide details about the product and borrowers represented. We detail additional data
used for cost estimates and exchange rate calculations. We then discuss the borrower survey.
Data for the project come from 3,705 loans at an Islamic bank in Pakistan. The bank
has national coverage, but borrowers with our loan are concentrated in the major cities
of Islamabad, Karachi, and Lahore. The sample includes all monthly repayments from

7
November 2003–June 2018, totaling over 275,000 observations. We observe each loan from
the start of repayment, but individual loans in the sample may be ongoing.
All borrowers in the sample have taken a diminishing musharakah, which is a popular
loan product for housing mortgages in Pakistan and other Islamic countries.2 In this model,
a fixed asset like a house or car is divided into equity units and the initial ownership is
divided between units owned by the consumer and units owned by the bank. Each month,
the consumer purchases one unit of equity back from the bank, and the bank charges a rental
fee on the remaining units it owns. These rental payments operate similar to interest, but
the structure allows the bank to capture profits without violating Islamic tenets prohibiting
the collection of interest (Obaidullah 2005).3
For consumers, the loan’s borrowing cost and repayment structure are similar to other
non-Islamic interest-bearing loans on the market, with one important caveat: diminishing
musharakah borrowers can also decide to purchase additional equity units each month to
shave months off the end of their loan tenure. While some other loans allow early pay-
ments (for example, refinancing a loan or making a large balloon payment), the diminishing
musharakah model makes it much easier to make these payments on a regular basis. Bor-
rowers need only to state the number of additional units (i.e. months of principal) they
wish to purchase upon each repayment and pay a fee equal to 3% of the repaid principal.
The low transaction cost to making an early payment leads many borrowers to make one or
several early payments over the course of their loan, and these repayments provide an easily
quantifiable measure of the desire to reduce one’s debt burden.
Descriptive statistics for the sample population are shown in Table 1. The average
annual income of borrowers is PKR 5,851,114 ($68,274 USD), which is significantly higher
2
We use the term “loan” here and throughout the paper to denote money borrowed to finance an asset.
We note that this definition diverges from the Islamic finance interpretation, where a “loan” denotes money
borrowed on interest without the presence of an asset, which is prohibited. In this context, diminishing
musharakah is not considered a loan. Nonetheless, we choose to use this term to allow for interpretability
for a general audience.
3
Similar to the prior note, we may refer to rental rates as interest rates in this paper for general inter-
pretability, but they should not be considered interest rates in the Islamic finance sense.

8
than Pakistan’s average income per capita over the sample period of PKR 94,270 ($1,100
USD). Anecdotally, the bank describes the typical customer for this loan as substantially
wealthier, more educated, and more urban than the general population of Pakistan.
In addition to the overall sample, Table 1 provides statistics on subjects who do and
do not make early payments over the course of the sample period. We assess three subsets
of these groups: All borrowers, Scheduled Repaid borrowers whose original loan contracts
were scheduled to end within the sample period, and All Repaid borrowers who have fully
repaid their loans in the sample period, regardless of when they were expected to complete
repayments. Among early payers, the All Repaid group includes the full Scheduled Repaid
group in addition to people who completed their loans earlier than expected due to early
payments. These Repaid groups are important to later analysis because we can observe
repayment behavior throughout the entire loan tenure, whereas the All borrower group
includes some loans that remain open at the end of our sample period.
Analysis of observables uncovers few differences between early payers and non-early pay-
ers, but suggests that the Scheduled Repaid group diverges from the sample of All Borrowers.
In Table 1, early payers and non-early payers in the All groups appear similar along observ-
able characteristics like income, loan size, loan interest rate (determined by credit risk), and
contract tenure. Among early payers, borrowers in the Scheduled Repaid group have higher
incomes, smaller loans, and shorter contract tenures relative to the sample population. The
All Repaid group is more reflective of the All borrower group and includes a larger sample
size (762 observations relative to 135 in the Scheduled Repaid group).
To estimate foregone savings income from early payments, we incorporate interest rates on
lending and savings accounts provided by the bank. Data on savings interest rates by month
were not accessible prior to December 2009 or after December 2016. To estimate interest
rates for savings accounts in the early years in our sample, we utilize data on deposit interest
rates (excluding zero markup accounts and interbank borrowing) for all private banks, as
reported by the State Bank of Pakistan Statistics and Data Warehouse Department. We

9
calculate the ratio of our bank’s rates to average rates for all private banks in Pakistan from
December 2009 to December 2016. We then multiply this ratio by average interest rates for
all private banks for missing periods to estimate our bank’s rate.
Throughout the paper, we apply a fixed conversion rate of 85.7 Pakistan Rupees (PKR)
to 1 US Dollar (USD) to allow for consistent conversions between currencies. This conversion
reflects the average monthly PKR to USD exchange rate between January 2004 and June
2018 using data from the International Monetary Fund, accessed through the UN Monthly
Bulletin of Statistics Online. We note that due to high inflation, the exchange rate at the
end of our sample period was much higher (PKR 121 to $1 USD), but we believe using the
average rate allows for both consistency and a reliable representation of the costliness of
early payments in dollar terms that is more understandable to most readers.
In addition to repayment data, we conducted a survey with a sample of 63 borrowers
with a diminishing musharakah loan from our bank. The bank provided phone numbers for
a convenience sample of 100 current borrowers from one urban location. Research assistants
called borrowers on the list to invite them to participate in a short phone survey. Seven
additional people completed the survey but are excluded from our analysis because they
reported that they had not ever held a diminishing musharakah loan. Since 10% of the
responding sample (7 out of 70) did not have diminishing musharakah loans, we also exclude
10% of the non-respondents to calculate an adjusted response rate of 70% (63 out of 90). To
encourage honest responses, the survey was anonymous. Subjects were told that the survey
was for academic purposes only and their individual responses would not be shared with the
bank. Subjects were not compensated for participating.

3 Results

In this section, we establish key facts about early payment behavior among diminishing
musharakah borrowers using data from the bank sample.

10
 Fact 1 : Early payments are frequent and sizeable: at least 40% of borrowers make early
payments, and the average early payer repays 85 months early.

In Panel A of Table 2, we find that 40% of all borrowers make an early payment over
the observable period. This figure is likely an underestimate because some borrowers in the
sample have a portion of their loan outstanding and may make an early payment before the
loan ends. When we restrict to only borrowers were scheduled to fully repay their loans
within the sample period, 35% of subjects make early payments. This lower proportion may
be due to differences in the sample that was scheduled to repay by June 2018 relative to the
overall sample. Borrowers in this “Scheduled Repaid” group had much shorter loan tenures,
so may have felt their loan was reasonably close to full repayment without the need for an
accelerated timeline.
Table 3 provides additional details about the size and frequency of early payments. Across
the full sample, borrowers make 3.3 early payments over the course of their loan and repay
85 extra units early. Each early payment includes an average of 25.7 extra units, which is
equivalent to paying off around two years of the loan tenure with one payment. Among
early payers who have completely repaid their loans, borrowers make an average of 4.1 early
payments and repay 108 units early, which aligns with our prediction that the “All Repaid”
sample overweights early payers making large payments. Among early payers scheduled to
repay their loans by June 2018, subjects make 2.9 payments and repay 42 units early. This
finding also aligns with expectations because the “Scheduled Repaid” group have smaller
loans overall and fewer total units to repay.

Fact 2 : Early payments can be financially harmful, costing borrowers an average of PKR
354,155 ($4,132 USD) in the sample.

11
 To estimate the financial impact of early payments, we consider three contributing factors
for each payment:

1. The fee of 3% of the unit of principal charged for making an early payment.

2. Foregone savings income from repaying principal early rather than keeping capital in a
savings account until the end of the loan, using it to make the last month(s)’ payment
and keeping any accumulated interest income. We assume annual compounding.

3. Reduction in interest (rent) charged on the unit of principal from the time of purchase
until the last period of the loan.

The impact of each individual payment will depend on the interest rate on savings and
lending at the time of purchase, as well as the months remaining in the loan tenure. To
understand conceptually how these factors interact, we first conduct a simulation of early
payments made at different times and in different interest rate environments.
In Figure 1, we plot the financial cost to a hypothetical consumer from repaying one
month of principal (one “unit”) at each month in a 20 year loan tenure. The figure highlights
that early payments are most costly early in the loan tenure because the cost of foregone
savings is highest when savings income has the most time to compound. At average interest
rates, a single extra unit purchased at the start of the loan would result in cost equal to
PKR 52,965 ($605) over the course of the loan. In the third quarter of loan tenure, payments
are much less costly and may even be financially beneficial under certain interest rates. An
extra unit purchased in the 170th month of the loan would result in a gain of PKR 4,306
($50 USD). Early payments increase in costliness in the fourth quarter because there are few
savings on interest (rental) payments but the borrower is still charged the 3% early fee.
With the simulation as a conceptual foundation, we estimate the cost of actual early
payments made by borrowers in our sample. We employ the same estimation methodology
used in the simulation, but utilize interest rates on savings and lending available at our bank,
described more fully in Section 2. In Table 3, we show that the average extra unit purchased

12
by borrowers in our sample results in a cost of PKR 3,869 ($45 USD). In total, early payers
increase their cost of borrowing PKR 354,155 ($4,132 USD) on average by making early
payments.
When restricting to only the sample of early payers who were scheduled to repay by June
2018, the average impact of early payments is actually positive. Borrowers save an average
of PKR 1,636 ($19 USD) per unit repaid early, or PKR 15,191 ($177) over the course of their
loan. Since borrowers in the Scheduled Repaid group had much shorter loan tenures, this
finding is not surprising given how the simulation demonstrated that early payments made
within 10 years of full repayment are often weakly positive.
In the All Repaid sample, the cost per repaid unit is similar to the full sample: subjects
lose PKR 3,365 ($39 USD) per extra unit repaid. Since borrowers in this group likely pay
more units on average, the overall cost is higher than the full sample. Subjects in this group
increase their borrowing costs by an average of PKR 519,526 ($6,062 USD) by making early
payments, relative to keeping funds in savings and repaying on the normal timeline.
To contextualize the size of these payments, we convert them into months of principal
using the average loan size for our entire sample. In the full sample, subjects lose about
one-fifth of a principal installment for each unit of principal they repurchase early. Over
the course of the loan, borrowers lose 17.6 months of principal from early payments, which
shows that they could have saved almost a year and a half of principal installments if they
had kept money in savings and repaid on a normal schedule.
As highlighted in the simulation, the cost of each payment varies by the time remaining
until full repayment, since this will dictate how long funds would have to compound in a
savings account. In Figure 2 we display the cost of an average unit purchased by quarter of
loan tenure for the full sample. Similar to the simulation, payments are most costly at the
beginning of the loan and least costly around the third quarter of loan tenure. The average
unit re-purchased in the first quarter costs borrowers an average of PKR 5,967 ($70 USD),
while the average unit purchased in the third quarter costs an extra PKR 1,070 ($13 USD).

13
In the simulation, the third quarter payments had a weakly positive financial impact. The
negative average impact in the actual sample is likely due to difference in lending and savings
rates when payments were made as well as shorter loan tenures due to early payments. While
the simulation analyzed payments in each month of a 20-year loan, the loan tenure of an
early payer may end up being much shorter when they repay many units early. In Figure 2,
a payment captured in the third quarter average may be made in year 15 of a 20 year loan,
or in year 9 if a subject repaid their loan in 12 years. These changes in tenure would impact
the cost estimates by affecting the years remaining until full repayment.
We believe these calculations reflect conservative estimates of costs for two reasons. First,
our method of calculating foregone savings income (i.e. the opportunity cost of early pay-
ments) uses interest rates that are likely lower than those available to consumers. We use
interest rates on term deposit savings accounts available through our bank because borrow-
ers would face low transaction costs to open these accounts. It is probable, however, that
other savings products in the market offer higher returns. Subjects could also achieve even
greater returns by investing funds in the stock market. The opportunity cost of using money
for early payments is therefore likely higher than what we calculate here, but since these
alternate uses of funds would require additional assumptions about transaction costs and risk
tolerance, we prefer to use our more conservative estimate of returns from savings accounts
that are immediately accessible to our sample.
Second, our estimates do not account for fees associated with late payments. 430 sub-
jects who made early payments in our sample repaid an installment at least 10 days late at
some point in the loan, leading to small fees that may have been avoided by keeping funds
in savings to repay on the normal schedule.4 We do not account for these fees here because
we believe these late payments were unexpected and thus these costs did not factor into the
decision to make an early payment.

4
Note that these fees are sometimes waived in practice, but they are stated in the loan contract.

14
 Fact 3: Early payments are more common early in the loan tenure.

In Figure 3, we analyze the distribution of early payments in each quarter of loan tenure.
We limit the sample to only subjects who have repaid loans fully in the sample period so
we can observe all early payments they make throughout the loan tenure. In Panel A, we
show the distribution of payments for this entire fully repaid sample of 736 early payers. The
proportion of early payments declines with each quarter of loan tenure. 35% of payments
are made in the first quarter relative to 12% in the fourth quarter. In Panel B we restrict to
the 136 early payers who were scheduled to fully repay their loan prior to June 2018. The
results are broadly similar, though more evenly distributed across quarters: 31% of early
payments are made in the first quarter relative to 15% in the fourth quarter.
In Figure 4, we analyze the distribution of total units repaid in each period. The All Re-
paid sample resembles the distribution of early payments, but the Scheduled Repaid sample
skews further to the early quarters in the loan. This finding suggests that in the latter sam-
ple, early payments contain more units on average when purchased earlier in the loan tenure.

Fact 4: Few early payments are made to fully close out a loan.

One reason people might seek to repay early is to reduce the stress and transaction costs
from making monthly payments. Under this hypothesis, the benefit of early repayment arises
only (or at least primarily) upon full repayment. If desire to fully close accounts drives early
payments, we would expect to see many early payments made in the final installment, and
an increase in early payments as the end of the loan tenure approaches.
In Figure 5, we examine the number of early payments made in each month during the
last three years of loan tenure among all subjects who fully repay in the sample period. We
find that the rate of early payments declines in the last two years of repayment. There is
only a small spike in the final period, which represents payments made to fully close out the

15
loan. In Table 2 Panel B, we compare early payment rates at the end of the loan to rates
during periods earlier in the loan tenure. In the All Repaid sample, early payers are about
half as likely to make payments in the last three months relative to payments in the first
three months of the loan. 3% of early payments are made in the last three months, relative
to 6% in the first three months.

4 Model

In this section, we develop a model of consumer hedonics that attempts to explain the
repayment facts established in the prior section. We start with a model that assumes people
are not debt averse, but quickly establish that early payments cannot be rationalized under
this formulation. We next test a model that includes a parameter of debt account aversion,
i.e. a disutility from open debt accounts. We show that this model can explain the existence
of early payments, but not the timing we established in Facts 3 and 4. We test a third model
that includes a parameter for debt level aversion, i.e. a utility gain from the act of proactive
repayment. We show that this model can rationalize all the repayment facts established
above.

4.1 Model 1: No Debt Aversion

We begin with a three period model of utility. For simplicity, we assume linear utility, but
will explore implications of relaxing this assumption later. In each period, a person starts
with income w, which is assumed to be to be spent on consumption if not allocated to
repayment. The person makes a regular mortgage payment m in each period, which allows
them to fully repay a loan over three periods. Utility is discounted at rate δ = [0, 1] in each
period. Our base case utility in model 1 is:

u1 = w − m + δ(w − m) + δ 2 (w − m)

16
 The person faces the option to make an early payment in period 2, which would enable
them to fully repay their loan in period 2. This option carries an additional fee f . The
utility with an early payment in model 1 is:

u1e = w − m + δ(w − 2m − f ) + δ 2 (w)

Under these assumptions, a person would prefer making an early payment to the base
case utility if:

(−m − f ) > δ(−m)

If f > 0, i.e. making an early payment is financially costly, it is immediately apparent

that under this model, early payments would never be rational. If consumers discount future
consumption even slightly, they would never prefer to bring a painful payment forward one
period, let alone pay an extra fee to do so. This model therefore cannot explain Fact 1,
which showed many early payments made throughout the sample.

4.2 Model 2: Debt Account Aversion

We modify the model by adding debt account aversion, where a person faces rising disutility
with each open debt account. For each period where the loan has not been fully repaid, we
assume that the subject experiences disutility a. We assume payments are made at the start
of the period, so in the period where the person makes a final repayment, they will not suffer
a. The base case utility becomes:

u2 = w − m − a + δ(w − m − a) + δ 2 (w − m)

With an early payment, the person experiences utility:

u2e = w − m − a + δ(w − 2m − f ) + δ 2 (w)

17
 Comparing ub2 and ue2 , we see the person will prefer making an early payment if:

(−m − f ) > (−a) + δ(−m)

This formulation captures the notion that early payments may be rational when the disutil-
ity of having a debt account open for one more period outweighs the added expense of early
payment fees and the disutility of moving an installment forward one period. If we assume
the psychological costs of debt are high enough, we can thus reconcile Facts 1 and 2 (high
rate of early payments despite financial costs) with Model 2.

In order to explore the timing of early payments in Facts 3 and 4, we must introduce an
option for the person to make a one-unit early payment in period 1 instead of period 2. We
model the utility with the first period early payment as:

u2ep1 = w − 2m − a − f + δ(w − m) + δ 2 (w)

A subject will prefer to make an early payment in period 1 instead of period 2 if:

−m − f > δ(−m − f )

Making an early payment in period 1 moves a painful payment forward one period without
reducing stress from an open account because the loan is still repaid in period 2. Assuming
the person is at least somewhat present biased (so δ < 1), they would never choose to make
early payments before the last period of the loan. With this model, we cannot explain Fact
3 and 4, which found that few early payments were made in the last quarter or last month
of the loan tenure.

18
4.3 Model 3: Debt Level Aversion

We add to the model by allowing for debt level aversion, i.e. a boost in utility when an
early payment is made and debt is reduced, rather than only when the account is closed.
We capture this utility shock with parameter l in the period when an an early payment is
made. We continue to allow for debt account aversion through parameter a. The base case
has no early payments, so u3 is equivalent to u2 :

u3 = w − m − a + δ(w − m − a) + δ 2 (w − m)

If an early payment is made in period 2, utility will be:

u3e = w − m − a + δ(w − 2m − f + l) + δ 2 (w)

In this scenario, a subject will make an early payment if:

(−m − f + l) > (−a) + δ(−m)

Here, the person experiences positive utility from making the payment as well as earlier relief
from the stress of an open account, making early payments even more favorable. Facts 1
and 2 on the frequency and cost of early payments are thus also consistent with Model 3.
To test Fact 3 and 4 on the timing of payments in this new model, we derive utility for an
early payment made in period 1:

u3ep1 = w − 2m − a − f + l + δ(w − m) + δ 2 (w)

The person will prefer making an early payment in period 1 to period 2 if:

−m − f + l > δ(−m − f + l)

19
For this inequality to hold with discounting, it must be the case that l > m + f , which is
to say that the positive utility from making an early payment outweighs the negative utility
of an extra month’s principal and associated fees. If this is the case, the person would want
to move the early payment forward to period 1. We thus need to assume that the positive
utility from making an early payment is high in order for the model to align with Facts 3
and 4.

4.4 Implications

Though our model of debt level aversion seems consistent with the basic facts established
in the prior section, it generates several implications that warrant further exploration and
testing:

Implication 1: The utility increase from repaying early must be high.

The model suggests that people are driven to make early payments by both stress relief upon
full repayment and the positive psychological from the act of paying early. These impacts
must outweigh the extra financial costs and the pain of moving a payment forward, suggest-
ing they must be sizable.

Implication 2: Debt level aversion is a separate preference from debt account aversion.
In the model, the parameters for debt account aversion and debt level aversion are fully
separable, meaning a person could experience a high value for one and not the other. For
example, a person could find debt stressful and thus be account averse, but not find enjoy-
ment from the act of making an early payment, leading them to either not repay early or
only make early payments at the end of their loan.

Implication 3: Early payments will be less frequent in periods when income is lower.
While this implication does not apply under linear utility, if we modify our assumptions to

20
make utility a concave function of total disposable income each period, it is apparent that
extra payments will be less frequent in periods when income is lower.

5 Supporting Evidence

In this section, we provide supporting evidence for the model implications described in Sec-
tion 4. Where possible, we conduct analyses using the repayment data. We also reference
prior literature and incorporate evidence from a small survey of borrowers on their subjective
perceptions of debt and early payments.

Implication 1: The utility increase from repaying early must be high.

To explore the relation between early payments and psychological effects, we ask our
survey respondents a variety of questions about their perceptions of the mental impacts of
debt and early payments. In Table 6, we show that 87% of early payers agree that debt
causes stress relative to 67% of non-early payers. 100% of early payers say that debt makes
them uncomfortable, relative to 76% of non-early payers. The high rate of agreement across
both groups show that debt causes stress and discomfort a large majority of borrowers, and
the higher proportion of agreement among early payers shows a positive correlation between
negative psychological impacts of debt and demand for early payment.
Prior research also supports the idea that debt is stressful for many borrowers. In a quasi-
experimental study of debt holders in Singapore, closing a debt account was associated with
an 11% reduction in the likelihood of anxiety (Ong et al. 2019). Over-indebted individuals
in particular have worse psychological outcomes, with causality appearing to run in both
directions (Gathergood 2012). Collectively, this evidence suggests that the disutility of debt
parameter a in our model is likely to be large for many subjects.
To assess the magnitude of debt level aversion, we ask respondents whether early pay-
ments are associated with positive psychological impacts like pride and relief. We find 80%

21
of early payers say repaying early makes them proud, relative to 70% of non-early payers.
The connection here between debt repayment and pride may be especially strong because
the diminishing musharakah loans are used as housing mortgages, and home ownership may
be more important to borrowers than owning other assets. 80% of early payers say repay-
ing early would make them relieved, relative to 88% of non-early payers. Evidently, most
borrowers believe that there are positive psychological impacts from repaying debt early,
supporting a positive value for parameter l in the model. We don’t find agreement to be tied
strongly to whether a respondent has made an early payment in the past, suggesting that
perceived positive effects of early payments may not be the main determinant of who makes
an early payment and who does not.

Implication 2: Debt level aversion is a separate preference from debt account aversion.
In the modeling section, we argue that debt account aversion alone cannot explain the
repayment behavior we observe, specifically the timing of most early payments in the loan
tenure. To reconcile the model with our repayment facts, we introduce a parameter for debt
level aversion, i.e. a utility gain from the act of repaying proactively. This formulation, how-
ever, requires additional testing to understand whether debt account aversion (as captured
in prior studies) and debt level aversion are separable preferences and which factor appears
most related to the decision to make early payments.
To compare our results with earlier studies more directly, we measure debt account
aversion using a one period version of the debt management game from Amar et al. (2011)
and compare it to self-reported experience with early payments. We present our survey
sample of diminishing musharakah borrowers with six debt accounts with varying balances
and interest rates, which mirror the accounts in Amar et al. (2011) but presented as PKR
instead of USD. As in Round 1 of the debt management game, we tell subjects they have PKR
5,000 to spend on debt repayment this month. We ask how they would choose to allocate it
across the accounts. By design, the amount is large enough to close off the smallest balance

22
account but would only pay down a fraction of the account charging the highest interest. The
subject must choose between paying off a small account (perhaps maximizing satisfaction
or motivation) or paying off the highest interest account (minimizing cost) or selecting an
alternate strategy (such as paying off a fraction of all accounts).
In Table 4, we analyze the strategies of subjects in our survey and compare them to
self-reported early repayment behavior. We find that 40% of subjects choose to close out the
smallest account, compared to 12% of subjects in Round 1 of Amar et al. (2011). Respon-
dents who reported that they had made an early payment on their diminishing musharakah
loan were 31% less likely to take this approach, significant at the 0.01 level. Conversely,
early payers in the sample were 17% more likely than non-early payers to allocate the full
PKR 5,000 to the account with highest interest, though this result was not significant. Min-
imizing costs by paying down high interest loans would allow subjects to repay debt fastest
by leaving more money for principal repayment instead of interest. In this way, the strategy
preferred by early payers seems to align with the impact of early payments in that they
demonstrate a preference for speedy repayment.
These findings suggest that debt account aversion in earlier studies may be distinct from
the debt aversion we observe in our context through early payments, which supports the rep-
resentation of debt level aversion as a separate parameter. It is important to note, however,
that our study does not negate the findings of prior research on the importance of debt ac-
count aversion. In our context, subjects are charged a 3% fee on early payments, which was
highly salient to borrowers we interviewed and may suppress early payments at the end of a
loan when full relief seems proximate. Furthermore, prior studies have examined repayment
across several open accounts and found evidence of debt account aversion in people’s prefer-
ence to close small accounts first. In our study, we cannot see if borrowers have additional
debt accounts outside of a housing loan, which is likely their highest balance account. We
therefore would not expect to observe debt level aversion as captured by other studies. Our
findings thus do not suggest that debt account aversion does not exist, but rather that a

23
broader conception of debt aversion is needed to capture a preference for paying down debt
even when the account remains open.

Implication 3: Early payments will be less frequent in periods when income is lower.
Liquid resources are required to make an early payment, so availability of these funds
may help explain who makes early payments and when these payments are made. If short-
term liquidity appears highly predictive of early payments, it is possible that the demand for
repaying early is higher than it appears in the data because some borrowers would want to
repay faster but lack the capital to do so. Unfortunately, we lack access to savings account
information to know whether non-early payers possess the cash resources for early payments
but choose not to make them. We can, however, analyze the aggregate rate of early payments
during times when borrowers are likely to have more or less cash on hand to get a general
sense of the importance of short-term liquidity for predicting early payments.
We first analyze how the rate of early payments vary depending on the timing of monthly
installments. The day in the month when installments are due depends on the day that each
individual begins using the property tied to their loan.5 Due to unexpected delays in the legal
processes to purchase a house, there is wide variation between when the loan is approved and
when possession occurs (from one day to several months), so we argue that installment dates
are assigned quasi-randomly. Borrowers can then make payments at any point in the month
preceding their due date, but in practice most repay in the week leading up to the due date.
The quasi-exogenous assignment of installment dates is important because it may impact a
user’s liquidity at the time of payment. Prior research has found that financial shortfalls
are more frequent when there is a greater timing mismatch between receiving income and
making expenditures (Baugh and Wang 2018). Conversely, households spend more following
increases in liquid holdings (Olafsson and Pagel 2018). Most subjects in our sample are paid
on the first of the month, so people with installment dates towards the end of the month
5
Note that this feature is somewhat unique to the musharakah loan model. In a standard mortgage loan,
installments are typically due on the first of the month from when the sale closes.

24
will likely have less cash to spare when their payments are made.
In Table 7, we predict early payments using the week in the month when installments
are made. In Column 1, we find that a payment is 0.2 percentage points (11%) less likely to
include an early payment if the installment is due in the last week of the month (from day
22 onwards), significant at the 0.01 level. Payments are not significantly more or less likely
in any other week of the month. To better understand this finding, we predict different size
payments in Columns 2-3. We find the coefficient on fourth week is only significant when
predicting “small” early payments of six months or less, which represents the bottom 25th
percentile of early payments in terms of extra units purchased. This finding supports the
idea that early payments are influenced by short-term liquidity at the time of payment, since
liquid resources equivalent to more than six months of principal may not vary from the first
to the fourth week of the month.
We also test whether early payments are more common in months when liquidity is likely
to be higher for many borrowers. In Table 8, we predict early payments by month of year
and find that payments are most frequent in January-April. January is the most common
times for workers to receive raises, which can be substantial in Pakistan’s high inflation
environment. Between 2004 and 2018, Pakistan averaged 8.8% annual inflation, so workers
whose pay mirrors inflation might expect substantial nominal raises. Since expenses are
unlikely to grow as discretely, borrowers are likely more cash flush soon after receiving a
raise. The monthly frequencies of early payments thus support the idea that early payments
are more likely when short-term liquidity is higher.
These analyses provide suggestive evidence that liquid resources are an important pre-
condition to early payments. Because we cannot analyze short-term liquidity at the indi-
vidual level, it is impossible to say whether cash constraints prevent non-early payers from
expressing a similar underlying preference for faster repayment.

25
6 Alternative Explanations

In this section, we explore alternative explanations for our repayment facts. We rely on tests
with the repayment data where possible, in addition to findings from the borrower survey.

6.1 Religion

Given the religious framing of the diminishing musharakah product as well as the broader
importance of religion in Pakistan, it is important to test whether religious pressure explains
the high rate of early payment. We investigate the impact of religion in our data by exam-
ining the rate of early payments during two periods of religious salience: the holy month
of Ramadan and the month leading up to Hajj, the pilgrimage to Mecca that all Muslims
are expected to make at least once in their lifetimes. Ramadan is the most sacred month in
the Islamic calendar, and a period where religious observance is likely to be highest (Ahmad
et al. 2012). Approximately 97% of Pakistanis are Muslim (Hussain n.d.), so virtually all
borrowers in our sample would identify as Muslim and observe Ramadan. We note that liq-
uidity during this period may be lower due to celebratory expenditure, but if early payments
are motivated primarily by religious sentiment, we would still expect to see a higher rate of
payments made during this month. Hajj is observed by relatively fewer Muslims each year,
but participants are specifically directed to clear their debts before the pilgrimage begins.
We therefore examine early payments made in the month leading up to the start of Hajj.
In Table 5, we report the results of OLS and logit regressions for whether a monthly
repayment includes an early repayment, predicted on whether a payment occurs during
Ramadan or during the month before Hajj. Off a baseline of 1.8% of installments having
early payments, we find that payments are 0.5 percentage points (28%) less likely to include
an early payment during Ramadan and 0.2 percentage points (11%) less likely in the month
before Hajj. The logit results confirm the directionality and significance of these findings.
Importantly, the timing of Ramadan and Hajj are based on the Islamic lunar calendar,

26
which is 10-11 days shorter than the civil calendar. This divergence means that the timing
of these events cycle through the calendar year, so these findings are unlikely to represent
seasonal trends in payments. In addition to testing whether religion explains the timing of
early payments, we also explore whether differences in religious identity correlates with who
makes early payments. While all people in our sample are likely to identify as Muslim, they
may differ in their degree of observance and interpretation of religious directives around debt.
In our small survey sample, we ask borrowers whether they agree with the statement that
“People should avoid debt for religious reasons.” In Table 6, we do not find a meaningful
difference in agreement based on early payment experience: 57% of early payers agree with
this statement, compared to 55% of non-early payers.
From this suggestive evidence, we cannot conclude that religion is not an important factor
in influencing debt repayment. In interviews, we often found that Islamic rules about debt
were top of mind for borrowers and influenced overall attitudes towards loans and repayment.
These beliefs, however, have little power to explain who makes early payments and when
these payments occur.

6.2 Commitment Savings

The high rate of early payments established in Fact 1 may be driven by a desire to protect
earnings from one’s own superfluous spending, which prior research has referred to as so-
phisticated present bias (O’Donoghue and Rabin 1999). Rather than keeping cash accessible
where they might be tempted to spend it unnecessarily, people may prefer to place it in a
restricted account (sometimes called a commitment savings account) that would charge them
a penalty for accessing the funds too soon. In field tests of these accounts in the Philippines,
around 30% of clients preferred commitment savings to a standard savings account (Ashraf
et al. 2006). This preference was correlated with to lab-based measures of present bias.
Demand for commitment devices may also be driven by concerns that excess cash may be
subject to “kinship taxation” by friends and family who might ask to borrow it (Goldberg

27
2017, Squires 2017). Some borrowers may view debt repayment as an even more restricted
form of commitment savings because funds used to repay loans early can only be reclaimed
by taking out a new loan.
To test whether demand for commitment savings explains our early payment facts, we
first explore the availability of alternative commitment savings options for borrowers in
our sample. If borrowers have easy access to profitable, restricted accounts, they would
be unlikely to incur the costs documented in Fact 2 in order to use early payments as a
commitment savings vehicle. We visit the websites of the five largest banks in Pakistan and
find that all offer “term deposit” accounts that makes withdrawals more costly. Users commit
to keeping money in these accounts for a designated period and are typically charged a fee
and/or receive less interest (called “profit”) earnings if they withdraw early, which provides
the commitment device that prevents users from using funds for gratuitous purposes. At
the bank providing our data on diminishing musharakah loans, people receive 3.3 percentage
points higher interest earnings on average with a term deposit commitment of 5 years. As
part of their loan disbursal, all loan recipients are automatically given a basic savings account
which they can easily convert to a term deposit account, so the transaction costs to opening
a commitment savings account is low. If borrowers have a strong desire to protect savings
from their future selves or others, it would be much more profitable for them to keep money
in a fixed deposit account until the end of their loan rather than using that money to make
an early payment. In our small survey, we also test whether agreement with statements
tied to commitment savings correlates with self-reported early payments. We find that few
respondents believe they would be tempted spend savings unnecessarily and that there is
not a significant difference between early payers and non-early payers: 17% of early payers
agreed relative to 9% of early payers. Relatively more respondents said that friends and
family would ask to borrow some of the savings but early payers were less likely to agree:
47% of early payers agreed relative to 61% of non-early payers.
The availability of alternate accounts and consistent responses from early and non-early

28
payers does not rule out definitively that demand for commitment savings does not contribute
to a demand for early payments. Across this analysis as well as in borrower interviews, we
do not find evidence that concern about superfluous spending or kinship taxation are major
forces driving borrowers to make early payments.

6.3 Misunderstanding of Costs

We document in Fact 2 that early payments can be extremely costly, but it is unclear whether
borrowers understand or pay attention to these costs. Since some early payments do generate
small positive returns, people may believe that they are saving money by making early
payments, making the trade-off between psychological and financial costs more ambiguous.
To test for awareness that payments can be costly, we ask subjects in our sample whether
they believe that early payments are “responsible” even if they are charged a fee. Eighty-
seven percent of early payers agree compared to 88% of non-early payers, showing that most
borrowers believe early payments should be made despite financial consequences. Differences
in this belief does not appear to explain who makes payments and who does not.
We also ask respondents to identify the maximum fee that they would be willing to pay
to make an early payment in a hypothetical scenario. On average, early payers would accept
a fee of 6.0%, which is twice as large as the actual fee charged by the bank. Thirteen percent
of early payers would accept a fee of 25% or greater, which was the maximal bucket allowed
in the question.6
Finally, we explore whether borrowers may be learning about costs over time, leading
to the reduction in early payments over the loan tenure that we observe in Fact 3. Most
early payers do make multiple early payments over the course of their loan, suggesting that
learning from the first early payment does not typically suppress future early payments.
Furthermore, we observe in our cost simulation that payments made between the third and
6
Note that only 49 out of 63 survey respondents answered the survey question on fees they would accept.
While we informed respondents that their answers were for research purposes and would not be shared with
the bank, it is possible that some non-respondents believed we would use their answers to charge them higher
fees in the future and thus declined to respond.

29
fourth quarter of the loan have a small financial benefit. If subjects have learned more about
costs over the course of the loan, we would expect to see an increase in payments during
this period, which we do not observe. Last, anecdotal feedback from borrowers suggests
that learning may in fact lead to more early payments in the future because first-time early
payers observe how easy it is to make payments and they experience positive mental benefits,
making them more likely to do so in the future.
Collectively, these findings show that borrowers are generally aware that early payments
carry financial costs and many people are willing to pay them. While people seem to un-
derstand the directionality of the financial impact of early payments, we cannot determine
if they have accurately calculated the magnitude of these costs. We therefore cannot rule
out definitively that underestimating costs contributes somewhat to the high rate of early
payments that we observe.

7 Conclusion

Debt holders often face large financial and psychological costs to carrying debt. In this
project, we exploit an early payment feature of an Islamic finance loan to show that borrowers
may accept large financial penalties in exchange for psychological benefits. This project
highlights the magnitude of the mental burden on debtors, which seems to be especially high
early in the loan tenure. This burden seems reduced by proactively paying down debt, even
if an account is not fully closed.
From a policy perspective, our findings suggest that mechanisms to reduce the psycho-
logical burden of debt may often be valued more by consumers relative to efforts to reduce
borrowing costs. In the diminishing musharakah case for instance, consumers might benefit
from a designated savings account tied to their loan where they can store cash for future
payments but accumulate interest savings rather than making an early payment. This ac-
count could also provide a cushion for borrowers facing income shortfalls in the future to

30
avoid penalty fees. Creating a closer tie between the savings and debt accounts would en-
courage borrowers to group them into the same mental account. This might grant them the
positive sentiments of relief and pride through savings in the account, whereas previously
they needed to make an official early payment to gain these psychological benefits.
This project also highlights the need for further research on the mental burdens of debt.
Most prior work focuses on repayment across several low balance accounts, particularly credit
cards, and find a strong preference to fully close accounts. We study housing mortgages and
find benefits to faster repayment, even when an account is not fully closed. Further research
is needed to determine whether our findings generalize to other assets and loan sizes, as
well as other contexts. With access to data on consumers’ full debt portfolios and savings
accounts, future studies can compare whether debt account aversion in prior studies and
debt level aversion we document are related and better test how liquid resources impact
revealed preferences.

31
References

Ahmad, S., Goel, K., Maroof, K., Goel, P., Arif, M., Amir, M., & Abid, M. (2012). Psycho-
Social Behaviour and Health Benefits of Islamic Fasting During the Month of Ra-
madan. Journal of Community Medicine & Health Education, 02 (09). https://doi.
org/10.4172/2161-0711.1000178
Amar, M., Ariely, D., Ayal, S., Cryder, C. E., & Rick, S. I. (2011). Winning the Battle but
Losing the War: The Psychology of Debt Management [Publisher: SAGE Publications
Inc]. Journal of Marketing Research, 48 (SPL), S38–S50.
Ashraf, N., Karlan, D., & Yin, W. (2006). Tying Odysseus to the Mast: Evidence from a
Commitment Savings Product in the Philippines [Publisher: Oxford University Press].
The Quarterly Journal of Economics, 121 (2), 635–672.
Baugh, B., & Wang, J. (2018). When Is It Hard to Make Ends Meet? Working Paper.
Besharat, A., Carrillat, F. A., & Ladik, D. M. (2014). When Motivation Is Against Debtors’
Best Interest: The Illusion of Goal Progress in Credit Card Debt Repayment [Pub-
lisher: American Marketing Association]. Journal of Public Policy & Marketing, 33 (2),
143–158.
Gathergood, J. (2012). Debt and Depression: Causal Links and Social Norm Effects [Pub-
lisher: Wiley]. The Economic Journal, 122 (563), 1094–1114.
Gathergood, J., Mahoney, N., Stewart, N., & Weber, J. (2019). How Do Individuals Repay
Their Debt? The Balance-Matching Heuristic. American Economic Review, 109 (3),
844–875.
Goldberg, J. (2017). The effect of social pressure on expenditures in Malawi. Journal of
Economic Behavior & Organization, 143, 173–185.
Hussain, R. (n.d.). Pakistan. Retrieved August 1, 2020, from http://www.oxfordislamicstudies.
com/print/opr/t236/e0616

32
Kettle, K. L., Trudel, R., Blanchard, S. J., & Häubl, G. (2016). Repayment Concentration
and Consumer Motivation to Get Out of Debt. Journal of Consumer Research, 43 (3),
460–477.
Keys, B. J., Pope, D. G., & Pope, J. C. (2016). Failure to refinance. Journal of Financial
Economics, 122 (3), 482–499.
Mani, A., Mullainathan, S., Shafir, E., & Zhao, J. (2013). Poverty Impedes Cognitive Func-
tion. Science, 341 (6149), 976–980.
O’Donoghue, T., & Rabin, M. (1999). Doing It Now or Later. American Economic Review,
89 (1), 89.
Olafsson, A., & Pagel, M. (2018). The Liquid Hand-to-Mouth: Evidence from Personal Fi-
nance Management Software. The Review of Financial Studies, 31 (11), 4398–4446.
Ong, Q., Theseira, W., & Ng, I. Y. H. (2019). Reducing debt improves psychological function-
ing and changes decision-making in the poor. Proceedings of the National Academy
of Sciences, 116 (15), 7244–7249.
Ponce, A., Seira, E., & Zamarripa, G. (2017). Borrowing on the Wrong Credit Card? Evidence
from Mexico. American Economic Review, 107 (4), 1335–1361.
Richardson, T., Elliott, P., & Roberts, R. (2013). The relationship between personal unse-
cured debt and mental and physical health: A systematic review and meta-analysis.
Clinical Psychology Review, 33 (8), 1148–1162.
Squires, M. (2017). Kinship Taxation as an Impediment to Growth: Experimental Evidence
from Kenyan Microenterprises. Working Paper.
Stango, V., & Zinman, J. (2009). What Do Consumers Really Pay on Their Checking and
Credit Card Accounts? Explicit, Implicit, and Avoidable Costs. American Economic
Review, 99 (2), 424–429.
Sweet, E., Nandi, A., Adam, E. K., & McDade, T. W. (2013). The high price of debt:
Household financial debt and its impact on mental and physical health. Social Science
& Medicine, 91, 94–100.

33
 Table 1: Descriptive Statistics

All Non-Early Payers Early Payers

Borrowers All Sched/All Repaid All Sched Repaid All Repaid
(1) (2) (3) (4) (5) (6)
Annual Income (PKR) 5,851,114 5,827,782 4,642,998 5,886,350 6,233,590 5,651,787
(21,018,242) (16,108,844) (12,670,067) (26,784,077) (21,690,425) (30,505,467)
Log Annual Income 14.53 14.57 14.32 14.47 14.40 14.39
(1.33) (1.32) (1.26) (1.35) (1.34) (1.31)
Loan Size (PKR) 5,074,546 5,491,406 2,999,461 4,445,020 2,910,787 3,335,170
(9,253,708) (10,543,035) (4,183,875) (6,817,262) (3,844,202) (3,983,514)
Log Loan Size 14.82 14.85 14.34 14.78 14.40 14.59
(1.02) (1.06) (1.02) (0.95) (0.92) (0.88)
Loan Interest Rate 0.12 0.12 0.14 0.13 0.14 0.14
(0.03) (0.03) (0.02) (0.03) (0.02) (0.02)

34
Contract Tenure 16.38 16.21 8.21 16.63 8.59 15.68
(5.27) (5.63) (5.41) (4.67) (2.65) (5.07)
Observations 3,705 2,229 255 1,476 135 762
Notes: This table summarizes average background and loan characteristics for the borrower sample. Standard deviations are shown
in parentheses. Log annual income is the logged gross individual income before taxes in PKR for the primary borrower. Log loan
size is the logged size of the loan principal. Log loan size is the logged amount of loan principal in PKR. Loan Interest Rate is
the average interest rate charged to the individual over the sample period, which varies by individual based on credit background.
Contract Tenure is the loan tenure in years agreed in the original loan contract. Actual loan tenure is shorter if subjects make
early payments. Column 1 shows average characteristics for all borrowers in the sample. Columns 2-3 restricts to subjects who
did not make an early payment in the sample period. Column 2 includes all non-early payers, and Column 3 includes those who
were scheduled to fully repay their loan within the sample period, which is the same sample as all non-early payers who have fully
repaid. Column 4-6 restricts to borrowers who made at least one early payment in the sample period. Column 4 includes all early
payers, Column 5 includes those scheduled to fully repay their loan within the sample period, and Column 6 includes early payers
who fully repaid in the sample period, regardless of whether their original loan tenure fell within the sample period.
 Table 2: Proportion of Subjects Making Early Payments

All Scheduled Repaid All Repaid

Panel A: Full Sample (1) (2) (3)
Any point in sample period 0.40 0.35 0.72

Observations 3,705 390 1,062

Panel B: Early Payers Only
First quarter of loan - 0.64 0.61

Second quarter of loan - 0.47 0.58

Third quarter of loan - 0.36 0.36

Fourth quarter of loan - 0.19 0.20

First 3 months of loan - 0.19 0.19

First year of loan - 0.53 0.44

Last year of loan - 0.24 0.22

Last 3 months of loan - 0.05 0.09

Observations - 135 762

Notes: Values show the proportion of subjects in the column-designated sam-

ple who made an early payment in the row-designated time period. Panel A
includes all subjects, and Panel B restricts to subjects who make at least one
early payment over the sample period. Column 1 includes all subjects for the
panel sample. Column 2 restricts to subjects who have fully repaid and were
scheduled to fully repay their loans based on their original loan contract. Col-
umn 3 includes subjects who have fully repaid, including those whose original
contract tenure extended beyond the sample period. Quarters are defined by
realized loan tenure, rather than the loan tenure in the original contract.

35
 Table 3: Average Early Payment Number, Size, and Cost for Early Payers

All Early Payers Scheduled Repaid All Repaid

(1) (2) (3)
Number of early payments 3.3 2.9 4.1
(3.9) (3.7) (4.5)
Total units paid early 85.0 42.0 108.2
(56.9) (30.3) (56.7)
Cost per unit paid early (PKR) -3,869 1,636 -3,365
(16,646) (5,669) (14,288)
Cost per unit paid early (months) -0.2 0.1 -0.2
(0.8) (0.3) (0.7)
Total cost of early payments (PKR) -354,161 15,191 -519,537
(1,592,160) (499,490) (1,870,161)
Total cost of early payments (months) -17.6 0.8 -25.7
(78.9) (24.8) (92.7)
Observations 1,476 135 762

Notes: This table provides average early payment characteristics across three groups of early pay-
ers. Standard deviations are shown in parentheses. Column 1 includes all subjects who made at
least one early payment during the observed sample period, including subjects who have not fin-
ished repaying their loan and who may make additional early payments in the future. Column 2
includes only early payers who have fully repaid and were scheduled to fully repay their loans based
on their original loan contract. Column 3 includes all early payers who have fully repaid, including
those whose original contract tenure extended beyond the sample period. All statistics are calcu-
lated by first identifying the relevant figure for each subject and then averaging across individuals.
To calculate average cost per unit, for instance, we first calculate the total cost of early payments
to each individual divided by the total units that subject purchased. We then average this statistic
across all early payers to identify the average cost per unit. This approach weights statistics by in-
dividual, rather than by extra unit as would be calculated by dividing total costs across all subjects
by total units. Costs are reported first in PKR and then in average months of principal, which di-
vides costs by the average principal amount due in each installment, weighted again by individual.

36
 Table 4: Early Payments During Periods of Religious Salience

All Observations Early Payers Only

OLS Logit OLS Logit
(1) (2) (3) (4)
Month of Ramadan -0.005*** -0.295*** -0.011*** -0.287***
(0.001) (0.057) (0.002) (0.058)
Month before Hajj -0.002** -0.120** -0.006** -0.139**
(0.001) (0.056) (0.002) (0.056)
Constant 0.018*** -3.975*** 0.047*** -3.014***
(0.000) (0.016) (0.001) (0.016)
Observations 275,671 275,671 108,650 108,650

Notes: This table reports results for regressions on an indicator variable for
whether a monthly installment includes an early payment. Standard errors
are in parentheses (* p<.10 ** p<.05 *** p<.01). Column 1-2 use the entire
sample of monthly repayments between 2005 and June 2018. Columns 3-4
restrict the sample to only monthly installments from borrowers who make
at least one early payment within the sample period. Month of Ramadan
is an indicator variable for whether the installment was made during the Is-
lamic calendar month of Ramadan. Month before Hajj is an indicator vari-
able whether the installment was made during the 30 days preceeding the
start of Hajj, the annual pilgrammage to Mecca.

37
 Table 5: Early Payments By Week in Month of Installment

Any Small Medium Large

Early Payment Early Payment Early Payment Early Payment
(1) (2) (3) (4)
Second Week -0.0007 -0.0003 0.0002 -0.0006
(0.0008) (0.0005) (0.0006) (0.0004)
Third Week 0.0009 -0.0004 0.0009** 0.0003
(0.0006) (0.0003) (0.0004) (0.0003)
Fourth Week -0.0020*** -0.0014*** -0.0002 -0.0004
(0.0007) (0.0004) (0.0005) (0.0003)
Constant 0.0182*** 0.0056*** 0.0081*** 0.0045***
(0.0004) (0.0002) (0.0003) (0.0002)
Observations 275,671 275,671 275,671 275,671

Notes: This table reports results for OLS regressions where the dependent variable is an in-
dicator of whether a monthly installment included an early payment. Standard errors are
in parentheses (* p<.10 ** p<.05 *** p<.01). In Column 1, the indicator is for any early
payment. In Columns 2-4, the indicator is for an early payment of a specific size. Small
payments are less that 7 units (bottom 25th percentile), medium payments are between 7
and 36 units (25th-75th percentile), and large payments are greater than 36 units (top 25th
percentile). Covariates include indicator variables for the week in the month in which in-
stallments are regularly due, which is fixed and quasi-exogenous for each subject. First week
reflects payments due within the first seven days of the month. Fourth week includes pay-
ments due from day 22 until the end of the month. Includes all years and observations.

38
 Table 6: Subjective Agreement with Survey Statements

Early Payers Non-Early Payers Difference

(1) (2) (3)

Panel A: Psychological Impact

Debt causes me stress 0.87 0.67 0.200*
(0.106)
Debt makes me uncomfortable 1.00 0.76 0.242***
(0.080)
Repaying early makes me proud 0.80 0.70 0.103
(0.111)
Repaying early makes me relieved 0.80 0.88 -0.079
(0.093)

Panel B: Alternate Explanations for Early Payments

People should avoid debt for religious reasons 0.57 0.55 0.021
(0.127)
Early payments are responsible, despite fee 0.87 0.88 -0.012
(0.085)
My friends/family would ask for my savings 0.47 0.61 -0.139
(0.127)
I would be tempted to spend savings unnecessarily 0.17 0.09 0.076
(0.085)
Observations 30 33 63

Notes: This table summarizes the proportion of survey respondents who marked strongly agree or
agree for the listed statement. Column 1 includes subjects who self-reported that they had made an
early payment in the past on a musharakah loan. Column 2 includes subjects who have or have had
a musharakah loan but who did not report having made early payments. Column 3 shows the differ-
ence in proportion of early payers who agree relative to non-early payers. The standard deviation is
indicated in parentheses, and stars indicate significance where *** p<0.01, ** p<0.05, * p<0.1. p̂

39
 Table 7: Strategies in Debt Aversion Game

Close Smallest Account Pay Down Highest Interest

(1) (2)
Early Payer -0.312*** 0.170
(0.119) (0.126)
Observations 63 63

Notes: This table summarizes the results of an OLS regression of whether

early payers are more likely to prefer two types of debt averse strategies
to repaying multiple debts in a survey. Standard errors are in parentheses
(* p<.10 ** p<.05 *** p<.01). The dependent variable in Column 1 is
the strategy where subjects say they will first pay off the smallest balance
account, even though it does not have the lowest interest. This strategy
represents debt account aversion as defined by Amar et al 2011. The de-
pendent variable in Column 1 is the strategy where subjects prefer to pay
down the highest interest account first, which would allow them to repay
fastest. All survey respondents have held or currently hold a musharakah
loan. Early payers are respondents who self-reported that they had made
at least one early payment on their loan.

40
 Table 8: Early Payments By Month of Installment

Any Small Medium Large

Early Payment Early Payment Early Payment Early Payment
(1) (2) (3) (4)
Jan 0.003** 0.001 0.002* 0.001
(0.001) (0.001) (0.001) (0.001)
Feb 0.003** 0.001 0.001 0.001**
(0.001) (0.001) (0.001) (0.001)
Mar 0.003** 0.001 0.001 0.001
(0.001) (0.001) (0.001) (0.001)
Apr 0.004*** 0.000 0.002** 0.002**
(0.001) (0.001) (0.001) (0.001)
May 0.001 0.000 -0.000 0.001
(0.001) (0.001) (0.001) (0.001)
Jun -0.001 -0.000 -0.000 -0.001*
(0.001) (0.001) (0.001) (0.001)
Jul 0.002 0.001 0.001 0.000
(0.001) (0.001) (0.001) (0.001)
Aug 0.000 0.001 -0.000 0.000
(0.001) (0.001) (0.001) (0.001)
Sep 0.001 0.001 0.000 -0.000
(0.001) (0.001) (0.001) (0.001)
Oct 0.001 0.000 0.000 0.000
(0.001) (0.001) (0.001) (0.001)
Nov -0.000 -0.000 0.000 -0.000
(0.001) (0.001) (0.001) (0.001)
Constant 0.017*** 0.005*** 0.008*** 0.004***
(0.001) (0.000) (0.001) (0.000)
Observations 275,671 275,671 275,671 275,671

Notes: This table reports results for OLS regressions where the dependent variable is an in-
dicator of whether a monthly installment included an early payment.Standard errors are in
parentheses (* p<.10 ** p<.05 *** p<.01). In Column 1, the indicator is for any early pay-
ment. In Columns 2-4, the indicator is for an early payment of a specific size. Small payments
are less that 7 units (bottom 25th percentile), medium payments are between 7 and 36 units
(25th-75th percentile), and large payments are greater than 36 units (top 25th percentile).
Covariates include indicator variables for the41
month in which installments are regularly due,
which is fixed and quasi exogenous for each subject. Includes all years and observations.
Figures

Figure 1: Simulated Cost of One Extra Unit Repaid at Each Month of Loan Tenure

Notes: Cost reflects foregone savings interest, 3% added fee, and reduced interest on loan for
one additional unit purchased at each month in loan tenure. Uses average interest rates on
lending (12%) and savings (8%) throughout sample period and modal 20 year (240 month)
loan period.

42
 Figure 2: Cost of One Extra Unit Repaid by Quarter

0
Financial Impact of Extra Unit Repaid (PKR)

-1,070
-2,000

-2,273

-3,096
-6,000 -4,000

-5,967
1 2 3 4
Quarter of Actual Loan Tenure

Notes: Reflects the average cost per unit for all extra units (months) repaid in each quar-
ter. Quarter uses realized loan tenure, which in the case of early payments is shorter than
scheduled loan tenure at time of disbursal.

43
 Figure 3: Distribution of Early Payments by Quarter

.4
Panel A: All Repaid Panel B: Scheduled Repaid

0.35

0.31
0.31
.3

.3
0.27
Density of Early Payments

Density of Early Payments

0.27

0.22
.2

.2
0.15

0.12
.1

.1
0

0 1 2 3 4 0 1 2 3 4
Quarter of Actual Loan Tenure Quarter of Actual Loan Tenure

Notes: The figures reflect the distribution of early payments made by quarter of loan tenure.
Panel A includes observations from all subjects who completely repaid loans during the
sample period. Panel B includes observations from subjects who were scheduled to repay by
June 2018. Quarter uses realized loan tenure, which in the case of early payments is shorter
than scheduled loan tenure at time of disbursal.

44
 Figure 4: Distribution of Units Repaid Early by Quarter

Panel A: All Repaid Panel B: Scheduled Repaid

0.40
.4

.4
0.38

0.31 0.32
Quarter of Actual Loan Tenure

Quarter of Actual Loan Tenure

.3

.3
0.19
.2

.2
0.18

0.13

0.09
.1

.1
0

1 2 3 4 1 2 3 4

Notes: The figures reflect the distribution of total units repurchased in quarter of loan
tenure. Panel A includes observations from all subjects who completely repaid loans during
the sample period. Panel B includes observations from subjects who were scheduled to
repay by June 2018. Quarter uses realized loan tenure, which in the case of early payments
is shorter than scheduled loan tenure at time of disbursal.

45
 Figure 5: Early Payments in Final 36 Months

70 60
Number of Early Payments
40 50
30

40 30 20 10 0
Installments Remaining

Notes: The figure shows the number of early payments made by borrowers in the sample
in the final 36 months of loan tenure. The X-axis shows installments remaining until full
repayment, so payments made at 0 months are payments to fully close the loan. The sample
includes only subjects who have fully repaid their loans, i.e. subjects for whom the final 36
months are observable.

46