13  Borrowing

In this chapter I look at a series of facts about individual or household financial borrowing behaviour, examine where that behaviour is inconsistent with traditional economic explanations, and examine possible explanations that can account for the observed behaviour.

Under the economic framework we examined earlier, people borrow to smooth consumption. If income is very lumpy and a person has no liquid savings, it is possible to rationalise borrowing at any interest rate, provided there is no alternative product available with a superior rate.

We have already seen that people do not smooth their consumption over their lifetime. So what role does borrowing play?

Below I examine three lending products (credit cards, payday loans and mortgages), the role that these might play in customers’ lives, and some possible rationalisations for the pattern of use that we see.

13.1 Credit cards

Credit cards present three puzzles that a traditional economic framework has difficulty resolving.

  • People borrow far more on credit cards that you would expect than if they were exponentially discounters
  • People fail to choose credit cards with the lowest borrowing costs
  • People hold both high-cost credit card debt and liquid assets that earn low rates of return.

I then consider a simple example of how intertemporal discounting might affect credit card debt.

13.1.1 Excessive borrowing

Although it is possible to justify borrowing at any interest rate if income is sufficiently lumpy, the levels of observed credit card debt are hard to justify. In particular, the impatience required to justify the high levels of credit card debt does not reconcile with the patience required to justify the savings in illiquid assets such as housing and retirement accounts.

Present bias provides one possible explanation. As noted in the discussion in Section 12.3.1, illiquid savings are hard to access immediately, so the potential consumption of illiquid savings is substantially discounted by someone with high present bias. This enables the saving of illiquid assets. However, consumption using a credit card suffers no such discount. It can occur immediately. Meier and Sprenger (2010) found that more present biased individuals were more likely to have credit card debt and had higher levels of debt.

13.1.2 Poor card choices

The explanation of present bias is, however, incomplete, as demonstrated by another puzzle. People don’t choose the credit card with the lowest borrowing costs.

At least a part of this relates to customers being attracted by teaser rates, which they pay more attention to than the long-term rates they will end up paying.

Customers also exhibit poor understanding of exponential growth and how a credit card debt can compound over time. (Recall the compounding question that formed part of the financial literacy test.) Poor understanding of compounding can lead to an underestimation of the cost of high interest rates.

13.1.3 Co-holding debt and savings

People often hold both high-cost credit card debt and liquid assets that provide low rates of return. In one UK survey, 12% of households in the sample held an average of £3800 in revolving credit on which they incurred interest charges, while at the same time holding liquid assets that they could use to clear all of this debt (Gathergood and Weber (2014)).

One rational explanation for co-holding is that that some expenses must be paid by cash or direct debit, not credit card. This requirement means that funds must be available in these forms.

An alternative explanation is that co-holding is a self-control strategy. By reducing the amount of unused credit capacity, it may reduce future spending. (Note the use of mental accounts here.)

One shortfall with these explanations is that people who hold multiple cards do not minimise costs when using the cards they have. They pay little attention to relative interest rates when choosing which card to use. They don’t repay the card with highest interest rate first.

13.1.4 An example of intertemporal discounting and credit card debt

The following example illustrates how an exponential discounting agent and present-biased agent will consider payment of a credit card debt. Will they pay sooner and avoid interest, or will they delay the payment and incur extra costs? And will their decisions through time be consistent?

Consider an agent with utility function \(u(x_n)=x_n\) who receives income \(I\) in each of three periods, \(t=0,1,2\). They have a credit card with an interest free period and are considering whether to:

  1. Not use the credit card, which leads to stream of consumption equalling their income in each period: \(C_1=(0,I;1,I;2,I)\).

  2. Borrow to increase consumption by $\(X\) at \(t=0\) and pay the debt $\(X\) with no interest at \(t=1\), which leads to stream of consumption: \(C_2=(0,I+X;1,I-X;2,I)\).

  3. Borrow to increase consumption at \(t=0\) and pay the debt $\(X\) with 20% interest at \(t=2\), which leads to stream of consumption: \(C_3=(0,I+X;1,I;2,I-1.2X)\).

Exponential discounter

Suppose the agent is an exponential discounter who discounts each period by \(\delta\), with \(0<\delta<1\). Their utility in each period under each choice are as follows:

\(t=0\) \(t=1\) \(t=2\)
\(C_1\) \(I\) \(\delta I\) \(\delta^2I\)
\(C_2\) \(I+X\) \(\delta(I-X)\) \(\delta^2I\)
\(C_3\) \(I+X\) \(\delta I\) \(\delta^2(I-1.2X)\)

An expected utility maximiser will maximise the sum of the utilities across the three periods.

For any \(\delta\) less than one it can be seen that borrowing and paying at \(t=1\) gives higher utility than not borrowing:

\[\begin{align*} &U_0(C_1)=I+\delta I+\delta^2 I \\[6pt] &U_0(C_2)=I+X+\delta (I-X)+\delta^2 I \\[6pt] &U_0(C_1)<U_0(C_2)\text{ as }X-\delta X>0 \end{align*}\]

The agent prefers to borrow as they get the consumption today as opposed to equivalent consumption in the future when it is discounted.

The question then becomes which period they intend to pay for their borrowing. They will prefer to pay in \(t=1\) if:

\[\begin{align*} U_0(C_1)&>U_0(C_2) \\[6pt] I+X+\delta (I-X)+\delta^2 I&>I+X+\delta I+\delta^2 (I-1.2X) \\[6pt] \delta (-X)&>\delta^2 (-1.2X) \\[6pt] 1&<1.2\delta \\[6pt] \delta&>\frac{1}{1.2} \end{align*}\]

If \(\delta>\frac{1}{1.2}\) the agent will pay at \(t=1\). Higher \(\delta\) means the future has more weight than for low \(\delta\). The less discount that is applied, the more the interest payment degrades the utility of the agent.

What happens when the exponential discounter reaches \(t=1\) and reconsiders when they should pay? They will prefer to defer payment to \(t=2\) if:

\[\begin{align*} U_1(C_1)&>U_1(C_2) \\[6pt] I-X+\delta I&>I+\delta(I-1.2X) \\[6pt] (-X)&>\delta (-1.2X) \\[6pt] 1&<1.2\delta \\ \delta&>\frac{1}{1.2} \end{align*}\]

The condition is the same. The exponential discounting agent will not change their mind. They are time consistent. This is because the comparison between \(t=1\) and \(t=2\) always involve a single discount by a factor of \(\delta\) regardless of when they make this comparison.

Present-biased agent

Suppose the agent is a present-biased agent who discounts any delay by \(\beta\) and each period of delay by \(\delta\), with \(0<\beta<1\) and \(0<\delta<1\). Their utility in each period under each choice are as follows:

\(t=0\) \(t=1\) \(t=2\)
\(C_1\) \(I\) \(\beta\delta I\) \(\beta\delta^2I\)
\(C_2\) \(I+X\) \(\beta\delta(I-X)\) \(\beta\delta^2I\)
\(C_3\) \(I+X\) \(\beta\delta I\) \(\beta\delta^2(I-1.2X)\)

As for the exponential discounter, for any \(\beta\) or \(\delta\) less than one it can be seen that borrowing and paying at \(t=1\) gives higher utility than not borrowing:

\[\begin{align*} &U_0(C_1)=I+\beta\delta I+\beta\delta^2 I \\[6pt] &U_0(C_2)=I+X+\beta\delta (I-X)+\beta\delta^2 I \\[6pt] &U_0(C_1)<U_0(C_2)\text{ as }X-\beta\delta X>0 \end{align*}\]

The agent prefers to borrow as they get the consumption today as opposed to equivalent consumption in the future when it is discounted.

The question then becomes which period they intend to pay for their borrowing. They will prefer to pay in \(t=1\) if:

\[\begin{align*} U_0(C_1)&>U_0(C_2) \\[6pt] I+X+\beta\delta (I-X)+\beta\delta^2 I&>I+X+\beta\delta I+\beta\delta^2 (I-1.2X) \\[6pt] \beta\delta (-X)&>\beta\delta^2 (-1.2X) \\[6pt] 1&<1.2\delta \\[6pt] \delta&>\frac{1}{1.2} \end{align*}\]

If \(\delta>\frac{1}{1.2}\) the agent will pay at \(t=1\). Higher \(\delta\) means the future has more weight than for low \(\delta\). The less discount that is applied, the more than interest payment degrades the utility of the agent.

You will note that this is the same condition as for the exponential discounter. This is because the agent is comparing costs in two future times. As both are in the future, the discount for the first period of delay (\(\beta\)) is not relevant. It is only the discount of \(\delta\) between them that affects the decision.

What happens when the present-biased agent reaches \(t=1\) and can decide whether to stick with their intention to pay at \(t=1\) or leave the payment to \(t=2\)? They will prefer to defer payment to \(t=2\) if:

\[\begin{align*} U_1(C_1)&>U_1(C_2) \\[6pt] I-X+\beta\delta I&>I+\beta\delta(I-1.2X) \\[6pt] (-X)&>\beta\delta (-1.2X) \\[6pt] 1&<1.2\beta\delta \\[6pt] \beta\delta&>\frac{1}{1.2} \end{align*}\]

As both \(\beta\) and \(\delta\) are less than zero, this condition is less likely to be met than the original condition of \(\delta>\frac{1}{1.2}\). The present-biased agent is more likely to defer their payment to \(t=2\) if they reconsider their decision at \(t=1\). They may change their mind from their original decision at \(t=0\), which means that they are not time consistent.

The intuition is that when first considering when they will pay, both potential payment dates are in the future and subject to the discount \(\beta\) for the first period of delay. It is then only the long-term discount rate \(\delta\) that affects the time of their payment. When re-considering at \(t=1\), payment today is not discounted by \(\beta\), whereas the future payment is. Therefore, delaying becomes relatively more attractive.

This example points to the role of intertemporal discounting in the accumulation of credit card debt. Both exponential discounting and present bias can lead to borrowing and interested being incurred. It also suggest that someone with present bias may be more likely to defer payment (even if they did not initially intend to), accumulating debt and interest.

13.2 Payday loans

Relative to credit cards, payday loans charge a higher rate of interest, with short-term charges implying huge annual costs.

Payday lending has been subject to much regulatory and legislative action in Australia in recent years. Since 2012, payday loan interest and fees have been legislatively capped. The caps are:

  • Establishment fee of 20% of the amount borrowed
  • Maximum monthly fee of 4% of the amount borrowed
  • Default fees up to a maximum of double the amount you borrowed
  • Can also pass on government fees and charge missed payment fees and enforcement expenses

Even though capped, this structure can lead to very high interest rates, particularly when considered on an annual basis. Consider a one month loan with the establishment and monthly fee. That is effectively 24% interest for one month!

13.2.1 Harm to consumers

There is an active academic debate on whether payday loans are helpful or harmful.

On the evidence of harm, people tend to use payday loans even though less expensive options are available. Bertrand and Morse (2011) showed that better disclosure marginally reduces take-up, suggesting payday loan use is at least partly due to misunderstanding the terms or consequences of the loan. (We will tackle disclosure in more detail in chapters on production distribution and regulation.) Further, the debt burden created by payday loans can lead to a debt spiral that harms the ability to cover basic financial needs.

An important consideration, however, is the counterfactual of whether the harm would occur in the absence of the payday loans. Bhutta et al. (2016) found evidence that, when payday lending is banned, people shift to other forms of high-interest credit rather than shifting back to traditional credit instruments. This may suggest that constraints to payday lending are addressing the symptom rather than the cause.

13.2.2 Who uses payday lenders?

Payday loan use is linked to low self control and low financial literacy.

Gathergood (2012) examined payday loan use in a survey sample where self control was measured by self-reported agreement with statements such as “I am impulsive and tend to buy things even when I can’t really afford them.” He found a that those with low self-control were more likely to use payday loans, although there were various mechanisms by which this occurred. Low self-control people had more income shocks. They were more likely to have other sources of credit withdrawn. They had more unforeseen durable expenses. All of these could trigger a need for high-cost short-term credit.

You can think about the low self-control in terms of present bias. Payday lending attractive is presence of high \(\beta\); that is, a large discount for any delay. However, the variety of mechanisms by which payday loans are required suggests we require a richer story than high present bias.

As for credit cards, financial literacy may also play a role. Payday lender users score poorly on tests of financial literacy. Lusardi and Bassa Scheresberg (2013) found that those with high financial literacy (measured by answering each of the numeracy, inflation and diversification questions) were around 5 percentage points less likely to use a payday lender (20% compared to 25% across the full sample).

13.2.3 Watch

To research payday lenders and understand why people use them, Lisa Servon worked as a cashier. She described her experience in The Unbanking of America (2017). Here, she talks about her research.

13.3 Mortgages

The major source of household credit in Australia is the mortgage. Mortgages comprise over 90% of household credit!

Below we examine two features of the Australian mortgage market: the difficulty in comparing loans, and the “loyalty tax” paid by those who stay with their home loan lender.

13.3.1 Comparing loans

Australian banks tend to advertise a headline variable interest rate for their mortgage products. Yet almost 90% of customers of the big four banks receive a discount from that rate (Australian Competition and Consumer Commission (2020)). This can include advertised discounts that they receive when obtaining the loan, and discretionary discounts that are given during the application process or after disbursement of the loan. The ACCC found that, as at 31 October 2019, the average discount on the headline variable rate for standard owner-occupier loans was between 1.23% and 1.31% for each of the four major banks.

The advertisement of rates that are not the rate paid means that interest rate comparison is weakly informative when shopping for a loan. And people tend not to do much shopping around. For instance, ASIC research found that 38% of mortgage customers visited only one mortgage provider (be that a lender or broker, but typically their existing financial institution), with another 26% visiting only two (typically their existing financial provider plus on other).

Research in the United States has highlighted the costs of failing to search for the best rate. Gurun et al. (2016) found the difference between the 5th and 95th percentile adjustable rate mortgage interest rate within a geographic region was 3.1 percentage points, and that was after accounting for borrower and loan characteristics.

13.3.2 Punishing loyalty

The ACCC found that existing borrowers pay around 0.26% interest more for their loan than new customers (as at 30 September 2019). If the existing loan is more than five years old, they are paying 0.40% more than what big four bank new customers are paying. As an estimate of the associated costs, those customers of more than 5 years had loans averaging $200,000. If they refinanced, they could save around $850 in the first year. Given these customers typically have lower loan balances and the lender knows the reliability of their repayment history, this difference in rate is hard to justify on basis of pricing for risk.

13.3.3 Explaining these phenomena

Both rational and psychological arguments can be constructed for the failure of customers to shop around.

On the rational, search takes time and has a cost. The benefits of any improvement in interest rates need to outweigh those costs.

However, the scale of the differences in interest rates makes it hard to justify the failure to search without assuming an unreasonably high cost of search or value of the borrowers time. In particular, most long-term borrowers could likely receive some further discount by sending an email or making a phone call requesting a discount (possibly accompanied by a threat to leave). A minimal cost action can achieve large long-term gain, but is not taken.

Present bias provides one explanation as the costs of search are today, whereas the benefits are distant. The benefits of the search receive unduly low weight to a hyperbolic discounter. This is still somewhat an incomplete explanation, as some of the steps to gain lower rates are of such low cost it requires unrealistic levels of present bias.

Another explanation relates to attention and knowledge. A customer with a long-term mortgage may not have given any attention to their current rate relative to the rates they could achieve in the market. The opacity of advertised rates would further cloud their comparison even if they were to focus attention. They do not take the steps to seek a reduced rate because they do not realise it is an option, not because they have calculated the costs and benefits of their action.