The extent to which financial behaviour can explain over-indebtedness amongst New Zealand families
3. The role of financial behaviour
This section outlines the theoretical and empirical basis for the relationship we would expect to find in New Zealand between financial behaviour and over-indebtedness or problem debt. The suitability of LSS 2004 data to measure this relationship is then discussed, followed by regression results. (Confounding effects of factors like income are discussed in Section 4.)
3.1. Basic model
The basic model being tested is that the likelihood of families experiencing over-indebtedness is a function of their financial behaviour:
Prob(over-indebted) = Fn(financial behaviour)
or more specifically
Prob(over-indebted) = α + β financial behaviour + ε (1)
where α is the intercept, β is the slope coefficient on financial behaviour or the degree to which a change in financial behaviour is correlated with a change in the likelihood of being over-indebted, and ε is the error term or the change in the likelihood of being over-indebted that cannot be explained by financial behaviour.
3.2. Theory and evidence
The theory and evidence on the role of behaviour in financial decision-making is limited, but compelling.
There is increasing evidence, notably from the fields of psychology and neuroscience, that decision-making behaviour is not as ‘rational’ as economists like to assume (see for example Akerlof, 2001).
Financial decision-making is influenced by habits (basic budgeting approaches such as living within your means), heuristics (basic rules of thumb such as saving 10 percent of gross income), and coping mechanisms (strategies or behaviours adopted to avoid or encourage a particular course of action such as using a shopping list or avoiding certain shops to avoid overspending). The Centre for Policy Development in Australia has recently published a paper which provides a good discussion of why these habits, heuristics and coping mechanisms may or may not result in so-called rational behaviour (McAuley, 2008).
Psychologists, neuroscientists and ‘behavioural’ economists alike argue that these habits, heuristics and coping mechanisms are in turn influenced by a range of personality and environmental factors.
These were discussed more fully in the earlier Families Commission and Retirement Commission report (Legge & Heynes, 2008), but to briefly recap, personality factors include:
- locus of control (the degree to which people consider they are in control of their own life and actions)
- aspirations (the degree to which people form aspirations based on comparisons with others)
- self-control (the degree to which people are impulsive or do not stick to long-term goals).
Environmental factors include:
- context relativity (the degree to which people’s decisions are influenced by the context in which they are presented)
- shared experiences (the degree to which people’s decisions are influenced by the experiences of family and friends)
- family decision-making (the degree to which people’s decisions are influenced by family processes and priorities)
- consumer socialisation (the process by which people develop an understanding of the economic world)
- aggressive lending and advertising (the degree to which people’s decisions are directly influenced by the actions of others).
The research in this paper focuses on personality factors, not because they are in some way more important than environmental factors, but because of the LSS 2004’s limitations in properly examining the latter.
Again, to summarise the key findings in the two Commissions’ earlier report, having an external locus of control (a tendency to believe your life is outside of your control), having aspirations based on comparisons with others, or having poor self-control (a tendency to be impulsive) have all been found to be associated with spending rather than saving habits:
Lunt and Livingstone (1992) compared those who saved regularly with those who did not in the United Kingdom and found that savers had more internal locus of control than non-savers, while non-savers tended to be fatalistic. In general, savers believed in personal control over finances, in budgeting and in keeping things simple, whereas non-savers tended to make life more complicated and felt less in control.
Lunt and Livingstone (1992) also found that those in debt not only experienced pleasure in consumption but also expressed their social worth and social relations through consumption, buying presents for themselves and others as rewards or bribes. Debtors also tended to talk more about money with friends. The authors commented that being in debt appeared linked to socio-psychological participation in consumer culture more generally.
Pinto, Mansfield and Parente (2004) found that students who tended to carry forward large unpaid balances were thought to make impulse purchases and use their credit cards to buy more than they could afford. Although these students were aware of the downsides of their usage level, they appeared unable to regulate or modify their behaviour in using credit.
Perhaps surprisingly, Pinto et al’s study does not support previous studies showing that the psychological factors of self-esteem and locus of control were inversely related to shopping behaviour and credit card spending. Regardless of their type of credit card use, the students reported very high self-esteem and stronger internal locus of control. This suggests that there may not be a straightforward relationship between locus of control, aspirations and self-control.
In terms of self-control, neuroscientists have recently isolated the brain circuit involved in thinking twice and checking impulsive action (Brass & Haggard, 2007). This provides a sound basis for the ‘hyperbolic consumption model’ developed by behavioural economists to model self-control problems (or ‘irrational’ consumer behaviour). According to this model, ‘hyperbolic’ consumers are like their ‘exponential’ counterparts in that they prefer instant gratification over achieving long-run goals (in other words, they have high discount rates or prefer consumption in the short term). Unlike their exponential counterparts, however, hyperbolic consumers also have time-inconsistent preferences – that is, their preferences change depending on whether they are asked what trade-off they would make now or in the future. (See for example Angeletos, Laibson, Tobacman, Repetto & Weinberg, 2001.)
Even if the relationship between locus of control, aspirations and self-control is complex, the potential link between any of these behaviours and over-indebtedness is appealing from a policy perspective because of the potential to alter behaviour (for example through appropriate education programmes). The link has not been thoroughly examined in the literature, however. A major limitation is the availability of robust data on behaviour. The LSS 2004 data offer some scope to examine this.
3.3. Behaviour in the Living Standards Survey 2004
The LSS 2004 provides information on personal style[4] and insurance[5], both of which tell us about families’ financial practices, which in turn could provide some insight into their underlying behaviour.
Personal style
Figure 3 summarises ‘all of the time’ and ‘most of the time’ responses to the personal style questions. These responses were considered to best reflect families’ regular behaviour. This information suggests that the majority of families[6] report prudent financial practices: knowing they can afford something before they buy it, knowing where their money goes, reading statements and paying bills on time.
While these responses effectively represent different financial practices, they do not necessarily represent different underlying behaviours or personality traits. To simplify the analysis, the research in this report has focused on responses to the statements about saving and living from one pay to the next.[7] There are a few reasons for doing this.
From a policy perspective it is of interest whether families who save are less likely to experience financial strain – that is, whether saving behaviour is an important resilience factor. Interestingly, the proportion of families who save appears relatively low: between 30 and 40 percent. As already mentioned, there is some evidence that in terms of underlying behaviour or personality – income constraints aside or ceteris paribus – that those with internal locus of control are more likely to save than those with external locus of control (Lunt & Livingstone, 1992).
Figure 3: Proportion of families with different financial behaviour reported ‘most’ or ‘all’ of the time
Livings Standards Survey 2004 (n=varies, weighted by EFU=varies)
A related question is whether those who live from one pay to the next are more likely to experience financial strain – that is, whether spending behaviour is a significant vulnerability factor. It could also be argued that people who have low self-control (or are impulsive) are more likely to live from one pay to the next than those with high self-control – again income constraints aside or ceteris paribus.
Figure 4 illustrates these financial practices and potential underlying behaviours. It shows that while most of those who save do not live from one pay to the next (79 percent), only around half of those who do not save live from one pay to the next (53 percent).
Figure 4: Saving and spending behaviour
| Yes 2163 46% Internal LOC? |
 |
Live from one pay to the next |
|
Yes 454 21% (10% EFUs) Low self-control? |
|
‘Saver and spender’ Internal LOC and high self-control? |
||
|
No 1709 79% (37% EFUs) High self-control? |
|
‘Saver but not spender’ Internal LOC and low self-control? |
|||||
| Routinely save OR Save for goals |
|
|||||||
| No 2491 54% External LOC? |
|
Live from one pay to the next |
|
Yes 1318 53% (28% EFUs) Low self-control? |
|
‘Spender but not saver’ External LOC and low self-control? |
||
|
No 1173 47% (25% EFUs) High self-control? |
|
‘Neither spender or saver’ External LOC and high self-control? |
Somewhat simplistically then, it seems that being a saver (having internal locus of control) means that you are unlikely to be a spender (having low self-control); but that not being a saver (have external locus of control) does not mean that you are more likely to be a spender. It is interesting to test then which behavioural combination is most strongly associated with problem debt. Consideration was given to the two-way relationship with financial strain, as illustrated in Figure 5. It is clear from Figure 5 that none of the behaviours have a one-to-one relationship with financial strain, but that saving is a likely resilience factor (hence a negative relationship is expected with problem debt) and spending a vulnerability factor (hence a positive relationship is expected with problem debt).
Insurance
Having insurance would appear to represent fiscally prudent behaviour and having more insurance would represent even more fiscally prudent behaviour. It could be argued that people with internal locus of control are also more likely to get insurance – again controlling for income or ceteris paribus (being able to afford insurance could be a significant confounding factor and is discussed in Section 4).
The LSS 2004 provides good information on families’ insurance behaviour.[8] The variable created is a dummy or binary variable of whether a family has insurance or not. While this variable is being used as a proxy for a family having internal locus of control (ie being in control of their own financial fortunes), the negative relationship expected with financial strain is to some extent a given because purchasing insurance presumably insures a person against having financial strain.
3.4. Estimation
The regression model set out in equation (1) can be re-specified in a number of ways, depending on the data used to approximate financial behaviour:
Ln(problem debt) = α + β1 ‘saver’ + ε. (1a)
Ln(problem debt) = α + β1 ‘saver and spender’ + ε. (1b)
Ln(problem debt) = α + β1 ‘saver but not spender’ + ε. (1c)
Ln(problem debt) = α + β1 ‘spender’ + ε. (1d)
Ln(problem debt) = α + β1 ‘spender but not saver’ + ε. (1e)
Ln(problem debt) = α + β1 ‘neither saver nor spender’ + ε. (1f)
Ln(problem debt) = α + β1 ‘insurance’ + ε. (1g)
The financial behaviour variables are all dummies or binary variables (0=no, 1=yes).
The basic model being estimated is effectively a logit model, as the dependent variable (financial strain) is also a dummy or binary variable with values 0 or 1. In other words, the problem that the model is trying to solve is the probability or likelihood that a family experiences financial strain, or that the family gets a value of 1.
It should be noted that the sample data used in the regressions are unweighted, and families that ‘don’t know’ or ‘refuse’ information are omitted from regression analysis.
It should also be noted that finding a statistically significant relationship between the explanatory variable(s) and problem debt does not automatically suggest a causal relationship. The theory and data provide the strongest basis for asserting a causal relationship. The theory still appears to be unproven but the reported experience of financial strain had to be in the last 12 months, and in the case of personal style, ‘most’ or ‘all of the time’ responses were selected to reflect persistent rather than fleeting or transitory behaviour. A weak causal relationship is therefore being assumed.
3.5. Results
Figure 6 shows the results of a series of separate bivariate regressions on different definitions of problem debt using different specifications of financial behaviour. As expected, saving behaviour is negatively correlated with financial strain and spending is positively correlated. Interestingly, the largest marginal effect for the first definition of problem debt (financial strain only) was achieved from the behavioural variable ‘spender not saver’. On its own, this variable appears to account for 42 percent of the likelihood of experiencing financial strain. Theoretically, these families have external locus of control and low self-control.
Being a ‘saver’ on the other hand (and being a ‘saver but not a spender’, which few ‘savers’ actually are) also accounts for around 30 percent of the likelihood of not experiencing financial strain. Theoretically, these families have internal locus of control and high self-control.
The effect of having insurance is also interesting. Theoretically, families who purchase insurance have internal locus of control. Similar to the ‘spender not saver’ variable, it appears to account for 41 percent of the likelihood of not experiencing financial strain. Unlike the spending and saving behaviours, however, the size of the effect appears to increase the more tightly problem debt is defined.
It should of course be noted that these regressions are bivariate only and provide a useful indicator of the effect and power of the explanatory variables – they are not in themselves strong models of problem debt. In most of the regressions the percentages of correct predictions from the models barely improved on the naïve estimate – the proportion of ‘1s’ in the problem debt variables. Also, as the next section outlines, the effects are likely to be confounded by other variables.
To simplify multivariate analysis in the next section, only the behavioural variables ‘spender not saver’ and ‘insurance’ are used.
Note: each value represents the result of a separate bivariate regression
Statistical significance: *10%, **=5%, *** 1%
3.6. Summary
Overseas evidence suggests that financial pathways and outcomes are influenced by a range of personality and environmental variables. Having an external locus of control (a tendency to believe your life is outside of your control), having aspirations based on comparisons with others, or having poor self-control (a tendency to be impulsive) have all been found to be associated with spending rather than saving habits.
The LSS 2004 data on financial practices can arguably be used to approximate locus of control and self-control. For the purposes of this study, therefore, families who live from one pay to the next most or all of the time are assumed to be ‘spenders’ and have low self-control or be impulsive; whereas individuals who routinely save or save for goals most or all of the time are assumed to be ‘savers’ and have internal locus of control. Families who purchase insurance are also assumed to have internal locus of control.
The variables ‘spender not saver’ and ‘insurance’ in particular have been shown to have a strong, but opposite effect on the likelihood of a family experiencing problem debt – the former effect being positive and the latter being negative.
Footnotes
- [4]
- Q162, LSS 2004 Questionnaire [Return to reference]
- [5]
- Q112, LSS 2004 Questionnaire [Return to reference]
- [6]
- It should be noted that information on personal style only pertains to the respondent, not the partner. It is possible that the respondent has sound financial behaviour, but is not mainly responsible for financial decision-making in the family. [Return to reference]
- [7]
- Future analysis could examine other financial practices more closely. [Return to reference]
- [8]
- Q112, LSS 2004 Questionnaire. The type and amount of insurance purchased is not explored in this study, but could be looked at in future analysis. Unlike personal style, this information refers to the respondent and their partner. A stronger case could therefore be made that this variable in fact captures family locus of control. [Return to reference]