Normative influence on retirement savings decisions: Do people care what employers and the government want?

Abstract

The need for Australians to increase retirement savings has been widely promoted. Yet our understanding of the motivations of individuals to save at a higher rate remains sparse. This article reports the findings of a survey of superannuation fund members and their intentions to contribute more to superannuation and to manage their investment strategy. The article uses the theory of planned behaviour to focus on the important motivational influence of social norms. Formative research identified a number of influential social referents. Among identified referents, the study found that spouses appear to be the primary source of social influence for retirement savings decisions. The government and employers appear to exert little influence, and financial advisors and superannuation funds take up the middle ground of social influence. Possibilities for interventions designed to influence behaviour are discussed; however, conclusions are tempered by the fact that correspondence between intention and behaviour is not tested in the present research.

Key words:

• investment choice
• retirement savings
• social norms
• superannuation

One of the most pressing economic issues to face Australian society over the coming decades is how retirement incomes will be funded for an increasing proportion of retirees. It is estimated that by 2050 there will be 23.5% of the Australian population aged 65 and over, compared with 10.7% currently. Moreover, the proportion of working‐age people will decrease from 5 people of working age to support each Australian aged 65 years currently, to only 2.7 by 2050 (Commonwealth of Australia, 2010) One example of the ageing and health‐related economic impact of this is that government expenditure will increase from 22% of gross domestic product (GDP) in 2015–16 to 27% of GDP by 2049–50 (Commonwealth of Australia, 2010).

A range of opinions has been expressed on current retirement savings adequacy. The Investment and Financial Services Association (IFSA 2006) suggested a savings gap of $452 billion existed, meaning ‘current superannuation still falls well short of the benchmark needed to fund adequate living standards in retirement’ (IFSA, 2006, p. 5). Rothman (2007) has identified some improvements in retirement savings adequacies, which are projected to increase further, through the ‘Better Super’ reforms of the Australian government. Russell, Brooks, Nair, and Fredline (2006) have also identified successful privately sponsored savings programmes. Notwithstanding these improvements, it has been identified that more than 3 million Australians are below targeted retirement savings adequacy levels (Access Economics, 2008).

Approaches to inducing people to save more have perhaps understandably focused on direct ‘hip‐pocket’ incentives: eliminating tax on superannuation withdrawals after age 60; introducing generous retirement transition rules; and gentler taper rates for pension eligibility. Early indications are that these have been successful, but more so for those at or near retirement and those with higher incomes (Rothman, 2007), although a good deal more needs to be done to motivate people to save more for their retirement needs. Yet much past research regarding the psychological determinants of individual's retirement savings choices can be characterised as disparate in that it fails to be couched within any integrative theoretical framework. One outcome of this situation is a lack of opportunity to assess the relative importance of various determinants of retirement savings behaviour.

This research project sought to redress the lack of research with a primary aim to identify the relative importance of key behavioural determinants and to relate these to intervention possibilities applied to the individual, the workplace, and the general public. The theory of planned behaviour (Ajzen, 1988) was used for these purposes as it has been widely applied in past research and shown to be robust across diverse behavioural contexts. The practical objectives of the research centre on identifying intervention opportunities to influence individuals' behaviour to achieve higher retirement savings accumulations. This is achieved by examining the influence of social norms on retirement savings decisions and drawing implications for educators, policy makers, and professionals as well as identifying areas for further research.

THEORY OF PLANNED BEHAVIOUR

Superannuation has become central to the Australian retirement income system, with employers contributing a mandatory 9% of employee wages to a complying superannuation fund. The motivations of individuals to make contributions beyond the mandated employer‐level, together with individuals' motivations to manage the manner in which accumulated savings are invested, were the two behaviours of interest to the research.

The theory of planned behaviour is predicated on three variables found to adequately predict the intention to perform a given behaviour (Ajzen & Fishbein, 2000). These variables are one's attitude towards the behaviour, one's perception of social pressure as a consequence of the views and actions of significant others (subjective or social norms), and one's perceptions of control over performance of the behaviour (perceived behavioural control). Measurement of these constructs is performed directly, according to multi‐item scales, and indirectly, according to expectancy‐value formulations of underlying behavioural (attitudinal), normative, and control beliefs. Intention and perceived behavioural control together predict actual performance of the behaviour in question.

The theory of planned behaviour has explained, on average across various contexts, 39% of the variance in intention and 27% of the variance in behaviour (Ajzen & Fishbein, 2000). Substantial bodies of theory and research support the construct validity and predictive validity of the theory of planned behaviour (for a review see Connor & Armitage, 1998).

The theory of planned behaviour was expected to effectively predict intention and behaviour in the present retirement savings context. However, the theory does not presume that any of its predictors will necessarily have primacy. Rather, relative predictive importance will depend upon the specific behaviour and the behavioural context under investigation. The relative importance of predictor variables in, for example, the health behaviour or travel domains may be expected to differ from that in the retirement savings domain because each domain is characterised by substantively different attitudinal, normative, and control factor considerations.

Social norms

Although the present research made no prediction concerning the relative predictive importance of antecedent variables, the subjective (social) norm variable was expected to prove influential. The literature on the influence of normative factors (e.g., ‘if others are doing it or recommending it, perhaps I should do it too’) is limited. However, Duflo and Saez (2003) provided some expectation that normative factors might be influential in retirement savings behaviour prediction. Duflo and Saez randomly selected participants who were offered a cash incentive to attend a benefits information fair, and found that those who attended the fair were significantly more likely to enrol in the savings programme. However, they also found the savings choices of individuals who did not attend the fair mirrored the savings choices of their fair‐attending peers. Duflo and Saez suggested that relatively minor factors (peer effects) that do not directly affect the financial attractiveness of saving can have a significant impact on the formulation of saving plans.

Bailey, Nofsinger, and O'Neill (2004) explored the role of social norm effects on retirement savings decisions in a US experimental setting. Social norms were found to have direct effects on contribution amounts. Bailey et al. suggested the results were important because, upon first being hired, a new employee will typically be asked to make decisions about participation and contribution levels. The studies of Duflo and Saez (2003) and Bailey et al. were performed in a US context. Employees in Australia are typically not asked about participation in voluntary contribution schemes (Fielding, 2007), in part because a mandated contribution scheme already exists and further voluntary schemes carry an administrative cost burden. Yet, for reasons discussed above, there remains a need for many individuals in Australia to save more for their retirement. Thus, wider voluntary contributions scheme application and uptake remains an important endeavour in an Australian context.

METHOD

Research design

The research followed the recommendations of Ajzen (2002a) with a questionnaire designed to elicit self‐reported responses to questions needed for measurement of the intention, attitude, social norm, and perceived behavioural control variables for two focal behaviours: contributing extra to superannuation within the next 12 months and changing the investment strategy of superannuation contributions within the next 12 months. Subsequent examination of superannuation fund records will enable the assessment of correspondence between participants' intention to perform the behaviours of interest and actual performance of the behaviours. This will be the focus of future work. The analysis reported here is confined to the causal path from the theory of planned behaviour's predictor variables (attitude, subjective norm, and perceived behavioural control) to the intention variable.

Participants

Participants were randomly selected from four Australian superannuation funds. Of a total of 20,000 questionnaires distributed by mail in 2006 (5,000 for each fund), 2,339 (12%) questionnaires were returned. It was estimated that the 67 questions plus demographic data would take 30-min for participants to complete and return. The modest response rate was hence anticipated but raised the possibility of bias in the data (Moser & Kalton, 1972; Oppenheim, 1966). It was not possible to compare the demographic characteristics of survey respondents with those of non‐respondents because the funds did not supply demographic details of non‐respondents. However, the population of interest was the Australia working population. Inspection of labour force data reveals that the average worker‐age is 39 years, men comprise 54% of the work force, and average worker annual earnings is $43,654 (Commonwealth of Australia, 2006a). Table-1 shows the demographic characteristics of the questionnaire sample. Women were over‐represented in the sample relative to the overall Australian population, although it is reflective of the overall fund membership profile of the four funds. The middle‐aged were similarly over‐represented and average participant income was slightly lower than the population average. The opportunity to perform gender‐ and age‐based analyses of the data alleviated some concerns about over‐ or under‐representation of these demographic groupings. Nevertheless, the generalisability of some aspects of the results remains subject to qualification.

Table 1 Sample demographics

Males (n)	916	(39%)
Females (n)	1,423	(61%)
Couples : Singles (%)		73:27
Average age (years)	45
Average annual income	$55,000
Average household income	$87,000
Average household mortgage	$70,000
Average household net wealth	$404,000
Average superannuation savings balance	$190,000

Table-2 compares the Australian population (Commonwealth of Australia, 2006b) occupation profile with the sample. The most notable difference is the over‐representation of professionals and under‐representation of technicians, sales workers, and labourers. To the extent that the data were not analysed for intergroup differences relating to these demographics, the results are subject to qualification.

Table 2 Sample occupation classification

	Managers (%)	Professionals (%)	Technicians, trade workers (%)	Community and personal service workers (%)	Clerical and administration (%)	Sales workers (%)	Machinery operators and drivers (%)	Labourers (%)
Sample (n = 2,216)	9	58	6	9	11	1	2	4
Census	14	21	13	9	16	10	7	10

Source. Commonwealth of Australia (2006b).

Measures

The recommendations of Ajzen (2002a) were followed in questionnaire design. Several items served as indicators of the latent variables (attitude, subjective norm, and perceived behavioural control) modelled to predict the latent intention variable. Both direct and indirect measures were used to measure attitude, subjective norm, and perceived behavioural control, as discussed below.

Direct measures

The direct measure of the attitude towards the two target behaviours was assessed by means of five 7‐point unipolar evaluative semantic differential scales (Osgood, Suci, & Tannenbaum, 1957). The anchors of these scales, modelled after Ajzen (2002a), were: harmful–beneficial, unpleasant–pleasant, bad–good, worthless–valuable, unenjoyable–enjoyable and wrong–right. Cronbach's alpha coefficients for attitude towards contributing extra and managing investment strategy, respectively, were 0.85 and 0.88.

With respect to the direct measure of subjective or social norm, respondents were asked to indicate on 7‐point unipolar scales the extent to which they believe that most people who are important to them, or whose opinion they value, think that: they should not–should perform the target behaviours; would expect them to perform the behaviours (extremely unlikely–extremely likely); would disapprove–approve of them performing the behaviours; would–would not perform the behaviours themselves; and intend to perform the behaviours themselves (completely false–completely true). Cronbach's alpha coefficients for subjective norm towards contributing extra and managing investment strategy, respectively, were 0.80 and 0.78.

Four items modelled after Ajzen's (2002a) method directly assessed perceived control over the target behaviours. Again using 7‐point unipolar scales, respondents were asked whether performance of the two behaviours would be impossible–possible, whether, if the respondent wanted to, he or she could perform the behaviour (definitely false–definitely true), the respondent's perception of the degree of control possessed over performing the behaviour (no control–complete control), and whether performance of the behaviour was mostly up to the respondent (strongly disagree–strongly agree). Cronbach's alpha coefficients for perceived behavioural control towards contributing extra and managing investment strategy, respectively, were 0.78 and 0.76.

Finally, three items assessed intentions to perform each of the focal behaviours. Participants indicated the extent to which they intend to (extremely unlikely–extremely likely), will try to (definitely false–definitely true), and plan to (strongly disagree–strongly agree) perform each behaviour. Cronbach's alpha coefficients for intention towards contributing extra and managing investment strategy, respectively, were 0.92 and 0.86.

Indirect measures

Accessible beliefs are assumed to provide the cognitive and affective foundations for attitudes, subjective norms, and perceptions of behavioural control (Ajzen, 2002a). If this assumption is correct, then beliefs can be relied upon to obtain indirect, belief‐based measures of these constructs. Accessible behavioural beliefs are assumed to account for attitude towards the behaviour, accessible normative beliefs for subjective norm, and accessible control beliefs for perceived behavioural control.

Prior to the main survey, a formative survey was conducted with an independent sample (n = 49) to elicit modal behavioural and normative and control beliefs (modal beliefs being those beliefs most commonly held in the population). Because the focus of this article is on the influence of normative beliefs on intention, discussion of behavioural (attitudinal) and control beliefs elicited in formative research is herein excluded. Thus, the following describes construction of the indirect measure for subjective norm alone. However, an identical process is followed to compute the attitude and perceived behavioural control indirect (belief‐based) measures.

The formative survey revealed five referent groups, common to both behaviours of interest: spouse/partner; financial advisor; employer; government; and superannuation fund. For each of these five identified referent groups, 7‐point unipolar scales assessed normative belief strength and motivation to comply. For example, the statement ‘My employer thinks that I should make extra superannuation contributions in the next 12 months’ was rated on a 7‐point scale (unlikely–likely) to produce a measure of normative belief strength. Similarly, to assess the motivation to comply, respondents rated on 7‐point scales the extent to which they care (not at all or very much) about what each referent would want them to do about their superannuation arrangements. A measure, comprising normative belief strength and motivation to comply with respect to each normative belief (referent), offers a ‘snap shot’ of perceived normative pressures in a given population (Ajzen, 2002a). An overall indirect measure of subjective norm can be obtained by applying the expectancy‐value model calculated as SN = Σn_im_i (where SN = subjective norm; n_i = normative belief strength i; and m_i = motivation to comply i). Thus, to produce a belief‐based estimate of subjective norm, belief strength scores were multiplied by motivation to comply scores and the resulting products were summed across all normative beliefs.

Procedure

The survey questionnaire was designed to minimise participant response ordering effects and participant fatigue effects. Different items assessing a given construct were separated and presented in a non‐systematic order, interspersed with items for the other constructs. Additionally, the questions were divided into four equal sets and rotated across participants. Moreover, care was taken in the questionnaire to counterbalance high and low end points of scales in order to counteract possible response sets.

Theory of planned behaviour questionnaires have been used by many previous studies across diverse contexts, and questionnaire design tends to closely follow Ajzen's (2002a) recommended format. The survey was distributed by the four superannuation funds with a covering letter of support from the fund. Before data analysis, scale counterbalancing was reversed so that high‐score end points reflected positive intentions in all cases.

RESULTS

Results for all direct and indirect measures are presented in this section. The subjective norm variable proved the most influential predictor of intention in the present, retirement savings, context. Following a summary of the relative importance of attitude and perceived behavioural control, the remainder of this section elaborates findings concerning normative influence on intentions.

Relative predictive importance of behavioural antecedents

Tables-3 and 4 present means for direct measures of the four key theory of planned behaviour variables and correlations among these variables. Scores could range from 1 to 7, with scale high‐points representing favourable evaluations in all scales. It can be seen from Table-3 that respondents had, on average, favourable attitudes towards contributing extra to superannuation within the subsequent 12‐month period, they perceived moderate social pressures to do so, they had high confidence that they could achieve their behavioural goals if they were to be pursued, and they were somewhat moderate in their intention to try.

Table 3 Attitude, subjective norm, and perceived control towards contributing extra to superannuation

Variable	M	SD	Correlation
			A	SN	PBC	I
Attitude (A)	4.84	1.23	–
Subjective norm (SN)	4.20	1.34	0.59	–
Perceived behavioural control (PBC)	5.24	1.65	0.38	0.35	–
Intention (I)	4.10	2.16	0.61	0.62	0.57	–

Note. Mean scores are based on scales, with possible scores ranging from 1 to 7.

All correlations were statistically significant p < .01, n = 2,283.

Table 4 Attitude, subjective norm, and perceived control towards changing investment strategy

Variable	M	SD	Correlation
			A	SN	PBC	I
Attitude (A)	4.47	1.14	–
Subjective norm (SN)	3.83	1.20	0.59	–
Perceived behavioural control (PBC)	5.91	1.27	0.16	0.10	–
Intention (I)	3.66	1.80	0.61	0.63	0.19	–

Note. Mean scores are based on scales, with possible scores ranging from 1 to 7.

All correlations were significant, p < .01, n = 2,285.

By contrast, Table-4 shows that, compared with making extra contributions, participants were less positive in their attitude towards changing their superannuation investment strategy within the subsequent 12‐month period. This difference between mean scores for each behaviour was significant.¹ Compared with making extra contributions, respondents perceived social pressures to make a change in investment strategy as significantly lower, but they had greater confidence they could achieve their behavioural goals if they were to be pursued. However, compared with making extra contributions, participants were moderate, yet significantly less inclined, in their intention to try to change investment strategy.

In sum, respondents believed, for both behaviours, that they possessed high control over performance of the behaviours. Their attitudes towards performance of the behaviours were moderately favourable, but, on average, they were moderate in their cognitions concerning social pressure to perform the behaviours. Despite these moderate to high motivational underpinnings, participants were, on average, moderate in their intentions to try to perform the behaviours. However, they were more inclined to try to contribute extra to superannuation rather than to change their superannuation investment strategy.

Predictive significance

The explanatory power for both behaviours of interest compared quite favourably with previous theory of planned behaviour studies that typically account for 39% of the variance in intention and 27% of the variance in behaviour (Armitage & Connor, 2001). Table-5 presents results of a regression of intention on attitude, subjective norm, and perceived behavioural control. R‐squared values of the regression for the intention to contribute to extra to superannuation of 0.76, and 0.72 for changing superannuation investment strategy, compare favourably with previous applications of the theory. The results confirm that the theory of planned behaviour predicts the intention to perform both behaviours of interest very well. The most important predictor of intention for both behaviours of interest was subjective norm. Table-5 indicates that this was particularly so for change to superannuation investment strategy, where the influence of control factors was greatly subordinate to subjective norm and, to a lesser degree, to attitude. For extra contributions behaviour, perceived behavioural control was marginally subordinate to the influence of subjective norm, but the influence of perceived behavioural control was far greater when compared to its influence on the intention to change superannuation investment strategy. Attitude ranked third in importance among the three predictors of the intention to make extra contributions to superannuation.

Table 5 Behaviour intention regressions

Behaviour intention	Independent variable	Coefficient (standard error)
Contributing extra to superannuation	Attitude	0.30** (0.057)
(n = 2283)	Subjective norm	0.74** (0.075)
	Perceived behavioural control	0.71** (0.048)
Change superannuation investment strategy	Attitude	0.25** (0.043)
(n = 2285)	Subjective norm	0.91** (0.067)
	Perceived behavioural control	0.13** (0.024)

Note.**Significant, p < .01. Unstandardised coefficient from regression of Intention on Attitude, Subjective Norm, and Perceived Behavioural Control for each targeted behaviour. R² values for contributions to extra to superannuation 0.76, 0.72 for changing superannuation investment strategy.

Normative influence

As discussed, the subjective norm variable was found to be the most influential predictor of the intention to perform both behaviours of interest. This section examines subjective norms more closely using both the direct and indirect measures.

It was noted previously that, based on mean scores for the direct measures of subjective norm, respondents generally felt moderately positive to neutral social pressure to perform the target behaviours (see Tables-1 and 2) and social pressure was strongly correlated with respondent's intention. The indirect composite belief‐based measure of subjective norm was moderately correlated with the direct measure of subjective norm (r = 0.35 for contributing extra to superannuation and r = 0.39 for change to superannuation investment strategy). Inspection of the correlation between normative belief‐based measures and intention in Tables-6 and 7 reveals that, for extra superannuation contributions, social pressure was most strongly associated with the individual's spouse or partner, as well as the individual's financial advisor. According to mean scores, participants viewed the wishes of financial advisors more strongly than spouses/partners, but participants were less motivated to comply with financial advisors than they were with spouses/partners. This same relationship between belief strength and motivation to comply was also true for change to superannuation investment strategy. Overall, the influence on intention of spouse/partner was stronger than that of financial advisor for both behaviours of interest.

Table 6 Normative belief strength and motivation to comply with important referents to make extra super contributions

Referent	Belief strength (n)		Motivation to comply (m)		Correlation nimi with Intention
	M	SD	M	SD
The government	5.45	2.08	2.94	2.21	0.07
Your superannuation fund	5.07	2.02	3.79	2.25	0.11
Your financial advisor	5.02	1.91	4.75	2.08	0.22
Your spouse/partner	4.06	2.02	5.23	193	0.35
Your employer	3.12	1.89	2.50	1.81	0.06

Note. Normative belief strength and motivation to comply scored from 1 (low) to 7.

n_im_i = Normative Belief Strength × Motivation to Comply provides indirect measure of subjective norm.

Table 7 Normative belief strength and motivation to comply with important referents to change super investment strategy

Referent	Belief strength (n)		Motivation to comply (m)		Correlation nimi with Intention
	M	SD	M	SD
The government	3.90	2.21	2.95	2.28	0.14
Your superannuation fund	4.01	2.25	3.79	2.09	0.16
Your financial advisor	4.46	2.08	4.75	1.97	0.24
Your spouse/partner	3.65	1.93	5.23	1.97	0.33
Your employer	2.77	1.81	2.50	1.80	0.15

Note. Normative belief strength and motivation to comply scored from 1 (low) to 7.

n_im_i = Normative Belief Strength × Motivation to Comply provides indirect measure of subjective norm.

Table-6 indicates that participants held quite strong beliefs about the wishes of government and, to a lesser extent, the wishes of their superannuation fund for making extra contributions to superannuation. However, participants did not appear at all willing to comply with the wishes of either the government or their superannuation fund. Among all social referents, participants rated lowest the wishes of employers in terms of both belief strength and motivation to comply. Accordingly, belief‐based measures of subjective norm for employers, together with the government, correlated lowest with intention to contribute extra to superannuation. Table-7 indicates that, for making changes to superannuation investment strategy, belief strength scores for all referents were comparatively lower than equivalent scores for contributing extra to superannuation. However, because a generic scale for motivation to comply was used for both behaviours, mean scores were the same for both behaviours of interest. Among the referent expectancy‐value multiplicative measures for making changes to superannuation investment strategy, correlations of employer and government with intention were lowest. In particular, the employer was a referent group that participants believed was hardly interested in their superannuation strategy change aspirations.

DISCUSSION

Intervention possibilities

The survey results provide clear policy implications. When using the theory of planned behaviour to explore intervention possibilities, it is important to consider both the mean level of an independent variable and its importance as described by its regression coefficient. If a particular independent variable's average score and its regression coefficient are high, then little may be gained by focusing intervention efforts on what is already a strong source of motivation. Alternatively, according to Ajzen (2002a), if an independent variable's mean score is low and its regression coefficient is high, then there is room to move in terms of guiding mean scores upwards with the objective of consequent increase in the mean levels of the dependent variable (intention). The present research found mean scores for the subjective norm independent variable to be lowest among the three predictors (attitude, subjective norm, perceived behavioural control) of intention and its regression coefficient was highest among the predictors. Accordingly, there appears to be a very strong case for interventions that illuminate normative behaviour.

The uncoupling of referent scores into their separate expectancy (belief strength) and evaluation (motivation to comply) components provides further information about intervention possibilities. Participants believed most strongly that the government was keen to see them contributing extra to superannuation. However, respondents were hardly inclined to comply with the wishes of the government. Similarly, respondents believed that their superannuation fund and financial advisor had an interest in them making extra contributions. However, respondents were more motivated to comply with their financial advisor than with their superannuation fund.

The data reveal that there is room for positive change in the influence of superannuation funds and financial advisors. Determining the means by which change might be achieved was beyond the scope of the research but is a recommended focus of future research. Some preliminary suggestions emerged from the formative study. An apparent degree of mistrust in the minds of participants concerning perceived vested interests of funds and advisors was revealed. That is, some participants believed that funds and/or advisors include in their products and services dubious claims of performance and are less than fully transparent about fee structure. The Federal Government's declared focus on more readable product disclosure documents is well placed. It is also strongly in the interests of funds themselves to be more transparent in seeking to receive increased contribution flows.

Although participants were inclined to comply with the wishes of their partners, participants' perception that their partners would wish them to contribute extra to superannuation was neutral. Neutral perceptions about the wishes of partners may point to a need to ensure that, on matters of superannuation, partners are dealt with jointly rather than separately by their funds and advisors. Finally, participants believed that employers had little interest in them contributing extra to superannuation and, among all referents, participants were least inclined to comply with the wishes of employers. Thus, another tentative conclusion is that employers are in a position to significantly improve employee intentions to contribute extra should employers succeed in garnering greater persuasive influence.

Inspection of normative beliefs for investment strategy change revealed that, as might be expected when compared to extra contribution beliefs, respondents were not of such strong belief that any of the referent groups would expect them to change their superannuation investment strategy. However, the pattern of mean belief strength and motivation to comply scores was similar to that which applies to extra superannuation contributions. Thus, the previous comments concerning intervention possibilities apply equally to interventions related to changing superannuation investment strategy. However, presently it is not an easy proposition for an individual to obtain regular low‐cost personal financial‐investment guidance. Given the motivational importance of face‐to‐face referents (spouse and financial advisor), printed materials, web‐based information and other impersonal forms of delivery of information, and advice alone may not be sufficient to achieve advocated outcomes.

Limitations

The Theory of Planned Behaviour (TPB) predicts that changes in behaviour follow from intention and perceived behavioural control. These two variables are thought to be the most proximal determinants of behaviour; for example, an individual is more likely to change to an advocated behaviour if he or she (1) intends to, and (2) believes that it is not difficult to change. However, other variables have been shown to directly influence behaviour after controlling for the influence of intention and behavioural control. Two notable examples are past behaviour (e.g., Ajzen, 2002b) and implementation intention (e.g., Gollwitzer & Sheeran, 2009). In recent times, the implementation intentions variable has received much support as a mediator of the intention–behaviour relation (Gollwitzer, 1999). Thus, implementation intention has been depicted as the most proximal determinant of behaviour. On the other hand, Ajzen, Czasch, and Flood (2009) argue that implementation intentions serve merely to increase task commitment, with which behaviour is more likely to follow. The present article adopts the view that intention provides the motivational impetus for behaviour, and we agree with the view of Ajzen and Fishbein (2005), who argue that complex social behaviour is based on conscious, situation‐specific cognitions. Thus, we argue that interventions afforded by the TPB should target intention, as well as the beliefs (cognitions) that underpin intention's direct antecedents. This approach does not preclude roles for past behaviour, behavioural control, and implementation intentions in behaviour prediction. However, the approach provides a platform for interventions by identifying relative predictive power among the factors that influence intention. It provides a focus for intervention efforts where one would otherwise be difficult to determine. We argue that, despite extant endeavour designed to better understand intention–behaviour relations, the approach adopted in the present article remains necessary if interventions are to be designed for application in important decision‐making domains. The question of whether the approach is entirely sufficient for these purposes is the subject of continuing research endeavour.

The survey was conducted prior to the global financial crisis (GFC) and prior to superannuation rule changes. Both suggest limitations on any conclusions drawn. However, Gerrans (2010) found that, comparable to levels previously observed, only 7% of fund members made an investment change over a 3‐year period including the GFC. This suggests that the need for individuals to review their investment strategy remains unchanged after GFC. The most significant rule change since the survey has been a reduction in the concessional contributions cap as well as proposed changes in the Super System Review. However, the new cap will only impact 1.8% of individuals (Commonwealth of Australia, 2009) and no legislated changes have emerged from the Review's proposals. Superannuation rule changes have become a constant and have altered, at the margin, the attractiveness of superannuation. The most recent changes are in this vein. They have not removed the fundamental role of superannuation, or changed the increased responsibility that individuals now bear for their retirement savings. Therefore, the conclusions drawn in this article are not reduced by these recent events.

Future work

By mandating employer contributions to superannuation, the Australian government has exercised a paternalistic approach to the problem of retirement savings. Notwithstanding improvements in aggregate and average superannuation savings levels, a large proportion of Australians remains below targeted retirement savings adequacy levels. The research in this article examined the motivational antecedents of two key retirement savings behaviours: making extra voluntary contributions and changing investment strategy.

The current research has identified several opportunities for intervention. As well as promotional and educational programmes that aim to influence beliefs of large numbers of individuals in society, there may be much more that can be done. In this regard, the agency of stakeholders other than the government appears to be at issue. For example, one of the more striking findings of the research was the opportunity for employers, with the support of funds and advisors, to more widely promote voluntary schemes with respect to both behaviours of interest. The manner in which promotional programmes that aim to move retirement savings behaviours in the advocated direction are designed and delivered was not the subject of the research, but in view of the findings, remains a matter worthy of further study.

Social norms represent an important area for interventions designed to influence retirement savings behaviour. The present research adds to the findings of Duflo and Saez (2003) and Bailey et al. (2004) by assessing the importance of social norms relative to other behavioural determinants. The research goes further by attaching importance weightings to different referent groups, thereby enabling better focus for intervention efforts.

Correspondence between intention and actual performance of the focal behaviours was not tested in this research. Accordingly, a limitation to interpreting behaviour causation from the intention and perceived behavioural control variables alone is noted. We suggest that other variables, such as implementation planning, social support, and past behaviour, may also play a role in predicting behaviour in the present context. Parsing relative importance among a broader set of variables antecedent to behaviour remains a task to be taken up in future research.

ACKNOWLEDGEMENTS

The authors gratefully acknowledge the support of superannuation funds: GESB, UniSuper, STA (AustralianSuper), and HESTA in the conduct of this research.

Notes

1. Significance is at the 99% confidence level unless otherwise noted.