RDP 2022-06: Do Australian Households Borrow to Keep up with the Joneses? 5. Results and Discussion

5.1 Baseline results

In this section, I discuss the baseline results from estimating Equation (1). Table 2 presents the estimated effects of a 1 standard deviation increase in Gini coefficient on the accumulation of total debt and different types of debt.^[6] Looking at debt as a share of income allows me to assess the economic significance of the estimates and enables easy and robust interpretation of the heterogeneous effects across sub-populations with different levels of income and debt in later sections.^[7]

The estimated effect of local income inequality on total household debt is negligible. The effects on home debt and other property debt are not statistically significant and the magnitudes small relative to the standard size of a mortgage. While an 8 percentage point response seems relatively large, on average mortgage debt is equivalent to 100 per cent of income, and represents more than 300 per cent of income for the top 10 percentile of debt-to-income ratios.

The effect on non-mortgage debt appears to be mostly driven by investment debt, which is both statistically and economically significant. A 1 SD increase in the Gini coefficient is associated with a little more than a 3 percentage point increase in the ratio of investment debt to income. For context, this is equivalent to the mean ratio in the sample. Business debt also increases but the estimate is not statistically significant.

Among types of consumer debt, only car debt significantly increases as the local inequality rises, with the standardised increase being around 0.5 percentage points. This compares to a mean ratio of car debt to income of 2 per cent.

Overall, the statistically and economically significant positive effects of local income inequality on investment debt and car debt are suggestive of the credit demand channel.^[8] While accumulation of car debt is consistent with traditional status signalling and trying to close consumption gaps, the accumulation of investment debt suggests an alternative interpretation of ‘keeping up with the Joneses’. That is, as income inequality rises in the local area, households may try to close the income gap with their neighbours by taking out more debt for the purpose of non-residential investment.

Nevertheless, to better understand the mechanisms and economic implications, it is important to understand which households drive the results. For instance, households at the upper end of the income distribution may not have much incentive to keep up. Therefore, if they drive the average debt accumulation, that could suggest the credit supply channel rather than the credit demand channel. Also, there may be financial or macroeconomic stability concerns if lower-income households or those with liquidity and credit constraints exhibit such debt-taking behaviour, in contrast with middle- and higher-income households who have a better ability to weather shocks. Finally, life cycle and housing status can inform us about the priorities in demand for debt, while willingness to take financial risk potentially motivates demand for investment debt.

In the next sections, I assess the mechanisms and the economic implications of local income inequality by estimating the heterogeneous accumulation of investment and car debt across key demographic and financial groups with Equation (2).

5.2 Income and financial prosperity

Table 3 (Panel A) presents the effects of a 1 standard deviation increase in Gini coefficient on investment and car debt accumulation by households in each local income quartile. Their weighted average constitutes the estimated effect in Table 2. I find that households in the middle-to-upper end of the local income distribution (the third quartile) increase their investment debt by 12 percentage points as a share of income, driving much of the overall effect on investment debt. By contrast, accumulation of investment debt by the bottom and top quartiles are neither statistically nor economically significant. These results reinforce the role of the demand channel, as investment debt accumulation is driven by households who have both the means and the desire to keep up with their richer neighbours and close the income gap. The heterogeneity in accumulation of car debt follows similar but more subdued patterns. Importantly, in both cases the results suggest limited concern for macrofinancial stability as these households are likely to be in a better position to weather shocks.

It is important to reiterate that I define income quartiles based on the local income distribution, rather than the national distribution. As such, households in the third quartile may still have low incomes by national standards. Nevertheless, if I categorise households by their positions in the national income distribution the results are similar (Table B9), with the third quartile driving the response (the third quartile has incomes ranging between $77,000 and $120,000).

Another potential limitation of using local income ranking is that households do not necessarily know where they stand exactly in the local income distribution. While inequality in the region could be reasonably common knowledge, households may perceive their relative position differently from a mechanically constructed income ranking. To the extent that this self-perceived position drives their demand for debt, particularly one that helps boost their future earnings, it is important that we assess the heterogeneity along this dimension. HILDA Survey data enable this assessment by asking households to self-evaluate their prosperity given current needs and financial responsibilities, with answers ranging from ‘very poor’ to ‘prosperous’.^[9]

Table 3 (Panel B) shows that households who feel ‘reasonably comfortable’ have the largest increase in investment debt following an increase in local inequality, consistent with results observed for middle-income earners. Unsurprisingly, accumulation of car debt is notably stronger among households who identify as financially prosperous, perhaps indicating their desire to signal their social status via conspicuous consumption, despite being in the middle of the actual income distribution.

5.3 Life cycle and housing status

Next I look at the heterogeneity over the life cycle of the household heads. Table 4 (Panel A) presents the estimates where households are distinguished by the age of the household head: 20–39 years of age, 40–64 years of age and 65 years of age and older. I find a significantly stronger effect of local income inequality on the accumulation of both investment debt and car debt for middle-aged (40–64) households. By contrast, younger and older households are less likely to adjust their debt holdings following a change in local inequality. These results broadly align with the earlier results, given middle-aged households are likely to have higher incomes, more established savings and balance sheets (relative to younger households). They are also consistent with life-cycle theories of investing, with older households generally expected to be less risk-taking in their investment strategies.

Consistent with these findings, households with a mortgage, who are mostly middle-aged, take out relatively more investment debt following a change in income inequality (Table 4, Panel B). Outright home owners have a reasonably small response in terms of investment debt, consistent with earlier findings for older households (who are more likely to be outright home owners). And renters, who are more likely to be younger, barely change their behaviour. The heterogeneous effects on car debt are less evident.

5.4 Financial constraints and attitude

Financial constraints and financial behaviour are often associated with household debt-taking behaviour (Johnson and Li 2010; Kreiner, Lassen and Leth-Petersen 2019). In this section, I examine whether accumulation of debt following changes in local income inequality differs based on measures of financial constraints. This is important to look at as there are concerns that constrained households could overextend themselves in trying to keep up with their neighbours. To measure constraints, first I follow Kaplan, Violante and Weidner (2014) to define the liquidity constrained (hand-to-mouth) as households that either carry no liquid wealth or have borrowed up to their credit limit at the end of the pay period. Hand-to-mouth households make up 15 per cent of the sample and are mostly younger to middle-aged, lower income and renters. This is in contrast with the United States where more than one-third of households are hand-to-mouth (Aguiar, Bils and Boar 2020). I also look at whether households would have difficulty raising money for an emergency, which captures credit and liquidity constraints.^[10] In this section I also assess the role of financial attitudes by looking at household's willingness to take financial risk with their spare cash.^[11] This is interesting, as it provides another lens to look at the role of credit demand versus credit supply in driving the results.

Table 5 (Panel A) shows that households that are not liquidity constrained (that is, who are not hand-to-mouth) take out larger loans for investment purposes than their liquidity-constrained counterparts. Similarly, rising local income inequality substantially changes debt-taking behaviours of households who are able to raise emergency funds easily (Table 5, Panel B). These findings suggest that constrained households are unable or unwilling to take on more debt for investment purposes, and are generally consistent with the earlier results based on incomes.

The differential accumulation of investment debt by financial attitude is even more striking, though not surprising. It highlights that only households willing to take financial risk borrow more for investment, following an increase in local inequality (Table 5, Panel C). As non-residential investment is a high risk–high return venture, this result lends further support to the theory that households try to close the increasing income gap. Again, these findings are generally consistent with the earlier results on prosperity.

Results for car debt paint a similar picture, with households that are not liquidity or credit constrained and are willing to take financial risk borrowing much more for car purchases. That said, the evidence is somewhat weaker compared to investment debt.

Footnotes

SD of changes within SA3s over time, equivalent to around 0.02 index points. [6]

Results for debt accumulation in monetary terms are robust and can be found in Table B5. [7]

I examine the extensive margin by estimating the probability of holding debt across different types of debt. Results in Table B1 suggest that the extensive margin does not matter and debt accumulation is driven by the intensive margin. [8]

Question asked in the HILDA Survey: ‘Given your current needs and financial responsibilities, would you say that you and your family are …’ with six possible answers, including ‘prosperous’, ‘very comfortable’, ‘reasonably comfortable’, ‘just getting along’, ‘poor’, and ‘very poor’. [9]

Question in the HILDA Survey: ‘Suppose you had only one week to raise $2000 for an emergency. Which of the following best describes how hard it would be for you to get that money?’ Answers are grouped into ‘can raise easily’ if respondents answered ‘Could easily raise emergency funds’ and ‘cannot raise easily’ for the rest. [10]

Question in the HILDA Survey: ‘Which of the following statements comes closest to describing the amount of financial risk that you are willing to take with your spare cash? That is, cash used for savings or investment.’ Answers are grouped into ‘willing to take risk’ if respondents answered ‘average’ and above and ‘not willing to take risk’ for the rest. [11]