Understanding Default Behaviour in Workplace Pensions: Automatic Enrolment in the UK.

Abstract This article examines the retirement savings behaviour of twenty-five 30-40 years olds automatically enrolled into a workplace pension scheme. Using qualitative interviews, the paper explores the interaction between savings motivation and willingness to adhere to, or deviate from, the pension scheme defaults. Integrating insights from different savings paradigms, including sociological approaches and behavioural economics, the paper highlights how social motives drove willingness to accept enrolment defaults. Participants’ reactions to the contribution defaults were motivated by a complex combination of factors including anchoring effects, the salience of ageing, and emotional responses such as pride, uncertainty and loss aversion. The author’s main premise is that greater attention needs to be given to the interaction between subjective feelings about saving for retirement and pension scheme design.

The United Kingdom (UK) introduced automatic enrolment (AE) into workplace pension schemes in  and, since then, over nine million employees have been enrolled (DWP, ). The UK's state-run earnings-related pension scheme has been discontinued and the remaining new State Pension pays a flatrate pension aimed at poverty alleviation. Workplace pension schemes are designed to provide supplemental retirement income to the State Pension. Participation in workplace schemes is not compulsory: employees may opt out, and around one in ten do so (DWP, ). Employees also have the right to increase contributions from the minimum default, which is currently % (up from % in April ). The DWP reports that the minimum default is too low to provide a satisfactory retirement income for most people. But we know little about how and why workers decide to opt out of workplace pensions or make additional savings beyond the default.
AE takes advantage of inertia to maintain scheme membership. The emotional bias for continuitythe status quo biasacts to hinder change and discourages the auto-enrolled individual from opting out (Thaler and Sunstein, , Hershey et al., ). However, critically, the status quo bias also impedes changes to contribution rates (Beshears et al., ). The challenge may be in achieving retirement income objectives using a voluntary system that has its roots in inertia and the little-understood field of behavioural biases. No country has implemented automatic enrolment on such a large scale before. With AE, the UK has launched a social policy experiment that has implications for the long-term well-being of millions of people. It is essential that we develop a better understanding of how the UK public saves for retirement in the context of automatic enrolment. Only then can policy-makers develop a pension system best able to meet the public's retirement income aspirations.
Factors associated with low retirement savings include income (Hershey et al., ), age and gender (Foster, ), low levels of financial literacy (Atkinson and Messy, , Lusardi and Mitchell, ), decision-making complexity (van Schie et al., ), uncertainty (Haselton and Nettle, ), social influence (Huh et al., ), self-esteem (Tang and Baker, ), salience (Clark et al., ), myopic preferences (Thaler and Sunstein, ), and failure of saving intention realisation because of inertia (Rabinovich and Webley, ). These various explanations are not mutually exclusive and the relationship and relative influence of sociological and cognitive factors may vary according to circumstances. However, the very fact of being automatically enrolled, and the associated inertia, have the potential to reinforce some of the negative motivators. For example, the initial contribution rate can become the status quo reference point against which future behaviour is measured. So, if someone is struggling with the complexity of deciding on a suitable contribution rate, having been enrolled at a default rather than being required to choose a contribution level themselves, this may actually make implementing change more difficult (Van Schie et al., ). Although strong motivation overcomes defaults (Sunstein, ), we know little about how motives interact with the pension scheme architecture. To achieve optimal pension savings rates, we must first understand the antecedents of current behaviour (Tang and Baker, ).
This paper contributes to that objective by providing multidisciplinary insight from an exploratory study into workplace savings behaviour. It reports findings of the research question 'under what circumstances do middle-to-high earners adhere to, or deviate from, the enrolment and contribution default settings in their workplace pension scheme? ' The study on which the article is based was inspired by interdisciplinary theoretical approaches recognising the value of combining demographics, psychological constructs and behavioural variables in retirement savings models (Hershey et al., , Strauss, ). The research integrates insights from different savings paradigms including sociological approaches and behavioural economics. The significance of the study is that it uses individual qualitative interviews to highlight the different motives driving behavioural responses to membership and contribution defaults in workplace pension schemes. The author's main premise is that greater attention needs to be given to the interaction between subjective feelings about saving for retirement and pension scheme design.
The influence of the pension scheme architecture Employers often use matching contributions to increase workplace pension participation but matches have also been shown to influence contribution decision-making. The impact of matching contributions on savings behaviour is not uniform across income bands. At low-income levels, the match increases scheme participation but, at higher-income levels, it has a negative effect on savings because employees tend to limit contributions to the maximum match (Huberman et al., ). One reading of this behaviour is that the match is interpreted as an optimal contribution rate (Madrian and Shea, ). This match effect is problematic because the UK's state pension is flat-rate and higher earners need to save relatively more for retirement than lower earners to maintain post-retirement income parity (DWP, ). In the absence of employers' matches, most auto-enrolled individuals start saving at the initial contribution default rate (Madrian and Shea, ). Statistically significant numbers of savers adhere to default contribution levels even though, in the absence of autoenrolment, they might have opted for a different rate (Beshears et al., ). It is recognised that inertia can work against increasing contributions and savers may adhere to default contribution levels for long periods even if these are sub-optimal (Beshears et al., ). One explanation is that individuals subconsciously attach undue attention to the numerical value of the defaulta process known as anchoring. Anchoring is a cognitive bias where the individual attaches too much importance to the first piece of information that they acquire. It is particularly prevalent in scenarios characterised by uncertainty and involving numerical value assessment (Tversky and Kahneman, ).

Conscious and unconscious mental processes
Although there is significant heterogeneity in default sensitivity, certain individuals, such as the young, are more likely to adhere to contribution defaults (Brown et al., ). Decision-making skills vary over the life course (Bruin et al., ), but, given the evidence that cognitive biases affect novices and experts alike (Morewedge et al., ), it is unlikely that relative youth is the main cause of adherence to defaults. However, pension researchers have long recognised a link between age and low retirement savings (Foster, ). Income constraints explain some low savings, as does lack of financial capability (Foster, ; Lusardi and Mitchell, ) but other factors may be at work such as an unwillingness to think about pensions. Adherence to a low pension default could simply be related to a lack of motivation to save because pensions are not seen as relevant (Clark et al., ).
Present-day bias can explain the failure to save adequately for retirement. Individuals may gain greater present utility from not saving than perceived future utility from having saved (O'Donoghue and Rabin, ). A consequence of discounting future utility is that myopic individuals may postpone actions that, in the longer term, may prove to be utility enhancing (Pietrzak and Tokarz, ). One may intend saving more for retirement but procrastinate because of the material drop in present consumption that saving would entail. Auto-enrolment is effective in helping overcome myopic procrastination about pension scheme enrolment but, to date, little has been said about its impact post-enrolment.
Procrastinating about increasing savings rates could be a consequence of the cognitive costs of deviating from the default: the more complex the decision and the greater the effort required, the greater the likelihood of an individual opting for the default option (Bernheim et al., ). Increased complexity in the decision-making process has also been demonstrated to lead to increased inertia. Beshears et al. () claim that, although the individual may know that the default is sub-optimal and intend altering it, they may delay in doing so because of the complexity of calculating an optimal rate. Although retirement goal clarity and financial knowledge are explanations for some active investment decision-making (Lusardi and Mitchell, ) insufficient cognitive effort alone does not explain all default effects (Smith et al., ). Arguably, default adherence is not a dysfunctional cognitive bias as such, but a rational response to the complexity of an uncertain world where there is insufficient information to make a decision (Huh et al., , Haselton and Nettle, ).
Increasingly, attention is being directed to the role of emotions on financial decision-making (Lerner et al., ). Incidental emotions experienced at the time of the decision can influence the evaluation of outcomes causing people to behave in ways that may be inconsistent with their long-term preferences (Loewenstein and Lerner, ). Arguably, incidental emotions may explain time-critical decision-making such as opting out of a pension scheme following enrolment. Integral affect, for example, is emotion that derives from thinking about the decision to be made (Loewenstein and Lerner, ). Anticipating future feelings, such as regret or satisfaction, may be relevant to financial decisions involving greater reflection, such as increasing contributions postenrolment. Including affect in reactions to pension scheme default architecture may prove a fruitful avenue of research.
There is merit, too, in moving beyond describing default responses as decontextualized cognitive phenomena and examining the normative context of pension schemes (Strauss, ). Huh at al. () argue that the social context can have a similar effect to actual defaults as observed choices become 'choice defaults'. One view is that there are sociological explanations for some default behaviour, perhaps linked to norm adherence in the workplace (Robertson-Rose, ; Duflo and Saez, ). It may be that the environment that individuals find themselves in alters cognitive responses. Robertson-Rose () identifies short-term employment contracts as a deterrent to pension scheme enrolment and thinking about contribution adequacy. This viewpoint is supported by Ambrus et al. () who find that subjects with constant expected income show less present-bias than subjects with non-constant income expectations. To advance our understanding in this area, it would be helpful to include subjective construal in the investigative framework and consider the motivations that individuals themselves feel are responsible for their behaviour.

Theoretical approach
The existing literature contributes to our understanding that employees may see the default as a starting point from which to make complex investment decisions. But we should not assume that all pension default adherence is a product of cognitive biases such as myopia and procrastination, or that all default deviation is a consequence of rational utility maximisation. To develop a fuller understanding of individual responses to default settings in a real-world environment, we need to examine the problem from an alternative epistemological perspective that recognises the importance of sociological factors. To that end, qualitative research is a valuable tool that can help explore the subliminal motivations for behaviour. A skilled interviewer can disentangle the complex interactions that underlie behaviour and can pursue a wider-ranging investigation than could readily be achieved by quantitative methods alone.
The research project underpinning this paper sought to advance our understanding of retirement savings behaviour that is not comprehensively supported by existing theoretical frameworks. To do so, it used an alternative investigative lens. The researcher adopted an exploratory approach and the exploratory nature of this research is of considerable importance. The paper does not seek to make statistical claims about the relationship between the research participants and their default adherence or deviation. The aim is to present qualitative empirical data which can be used to develop hypotheses and to illuminate directions for further study. The exploratory intentions of the study should temper the reader's response to the data. One needs to be clear about the ontological position of critical realism that underlies the research. Critical realists believe that reality exists but that our understanding of it can only ever be imperfect and probabilistic because all observation is fallible and all theory is revisable (Archer et al., ). With that in mind, this article adopts a qualitative approach situated in a UK AE context and sets out to answer the following research question: 'Under what circumstances do individuals on medium-to-high incomes adhere to, or deviate from, the enrolment and contribution default settings in their workplace pension scheme?'

Methodological approach
The research was designed as an embedded case study involving twenty-five employees of a single large publicly traded UK company. All participants had access to the same DC workplace pension scheme and, although work locations were spread throughout the UK, they all had access to identical comprehensive on-line pension scheme information (see Robertson-Rose, ). This similarity in pension provision enabled the researcher to control for the confounding effects of the pension scheme architecture.
Interviews took place just prior to the company introducing legislated AE in . But, critically, for the previous seven years, the employer had been voluntarily automatically enrolling new employees into its DC pension scheme. Enrolment was at a default contribution rate of % (subsequent contribution increases were freely permitted and there was no upper contribution level). The employer matched employee contributions up to a maximum of %. The company also awarded a loyalty bonus of an additional % after five years' service and a subsequent % loyalty bonus after ten years of service. Thus, a long-term employee contributing at % would receive an employer contribution of %.
The embedded cases were purposively selected based upon Mill's Method of Difference. The technique selects cases which are similar but where there is variation in outcome (the dependent variable). This approach allows the researcher to control for variables and facilitates cross-comparison between cases. The selection criteria for similarity was based on a theoretical understanding of the socioeconomic explanations for differences in retirement savings behaviour. Similarity was operationalised by age, income and education. Controlling for these variables facilitated the search for other explanations for pension contribution variation.
The participants were invited to participate in the study having been identified from responses to a researcher-designed company-wide workplace survey (N = ) (see Robertson-Rose, ). In addition to socio-demographic details, the survey collected information about propensity to save, present-day bias, assets and investment spread.
The participants were all on middle-to-high incomes. Income is judged to be an important determinant of participation and contributions into workplace pension schemes. It directly influences retirement savings behaviour and indirectly influences psychological factors such as future-time perspectives (Ambrus et al., , Hershey et al., ). Higher earners tend to have greater flexibility regarding their financial decisions and, in theory, the choices that they make about their contributions into workplace pensions reflect their priorities rather than absolute levels of affordability. The salary band £, to £, was selected to limit interaction with state-benefits or higher-rate tax liabilities that might otherwise influence behaviour.
Participants were aged between  and  years, capturing those with some pension provision but who were still a distance from retirement. Increased age positively correlates with saving for retirement (Foster, ). All participants had tertiary education. Education (and associated financial literacy) is positively correlated with asset accumulation (Lusardi and Mitchell, ). The survey included a financial literacy test based upon OECD guidance (Atkinson and Messy, ) and all participants obtained a score above the UK average for financial literacy. The sample included  females and  males (the gender distribution reflected the firm's gender balance).
One limitation of the study is that the purposive sampling limits the generalisability of the findings. As well as being higher-paid (the median gross UK salary for full-time employees at the time was £,), the participants were slightly younger than the average UK employee. They were also better educated and more financially literate than average. UK Higher education participation rates were around % when this cohort graduated (although are now higher). The findings relate only to the participants and the study cannot confirm whether the behaviours identified in this group are prevalent in other populations. That said, one can make analytical generalisations where there is duplication of behaviours between cases.
Despite the socio-economic similarities, there was considerable variation in enrolment and contribution patterns. Eighteen participants were automatically enrolled upon recruitment (two opted out but one subsequently re-joined). Eight longer-serving participants joined the firm before its introduction of AE in , of whom six actively enrolled, mostly following a period of non-membership. (In , following legislated AE, the three non-enrolled participants were subsequently auto-enrolled into the scheme). Eight individuals were contributing at %, ten at % and one each at %, %, % and % (see Appendix). Some of this enrolment and contribution variation may reflect differences in career histories and informational and social network effects which are discussed elsewhere (see Robertson-Rose, ).
Semi-structured interviews were held in six workplace cafes at locations across the UK and lasted approximately one hour. The prior survey data enabled the interviewer to engage in an informed conversation that built upon knowledge of participants' socio-economic background. Participants were asked to 'tell the story' of their experience of pensions and this enabled a deeper understanding of the motivations behind savings behaviours. Free conversation was interspersed with pre-prepared questions developed from pilot-study interviews, giving focus to the conversations and providing continuity that aided later comparative analysis. Interviews were recorded and transcribed. NVivo software was used to apply descriptive and pattern coding and to identify commonalities and differences. Coding and analysis were peer-reviewed. Evaluation involved the constant comparative method, in which themes identified in the newly acquired interview data was compared to the existing data. Identified themes were used to inform subsequent questioning. In this way, the data collection and analysis developed as an iterative process.

Research findings
Accepting membership defaults and avoiding regret A recurrent theme in the interviews was that saving into a pension was viewed as necessary. Participants frequently used terms such as 'you have to', 'you need to', 'you ought to' and stressed the importance of 'doing something' about preparing financially for retirement because 'It's important. You got to plan for the future' (D). It was not unusual for participants to claim that they would have joined the pension scheme even in the absence of AE. Scheme membership was often justified with reference to the approval of others-usually parents or work colleagues. Several participants linked saving for retirement with having children because 'it's the right thing to do' (B). Thus, the motivation to remain enrolled in the workplace pension scheme appeared to be being reinforced by social norms about 'responsible' behaviour.
The wish to put down a marker for retirement was not restricted to participants saving at the minimum defaulteven those who had increased contributions placed greater stress on the importance of membership than the fund values that higher contributions would generate. For several participants, such as Q, it was important that, in the future, they could justify their current savings behaviour both to themselves and others. Social and psychological motivations also appeared to override 'present-day bias' (O'Donoghue and Rabin, ). As part of the pre-interview fact-find, participants were asked two questions designed to detect present-day bias. These questions were: • If I had a choice of receiving £ today or £, in one year's time I would choose : : : ? • I would rather have a good standard of living today than save for retirement.
All the participants with present-day bias were enrolled in the pension scheme. Some stated that the future was low priority but that pension scheme membership was important. However, the emphasis was often upon the 'act of saving'of ensuring that the pension 'box was ticked' (G), rather than upon actively reflecting on the financial situation they would likely experience upon retirement. There appeared to be a psychological benefit to moving from a non-enrolled to an enrolled state, perhaps because being enrolled symbolised 'ideal' behaviour. But enrolment also removed some anxiety about not preparing for retirement. It appeared to reduce the perceived need to consider retirement outcomes more deeply (or at least for the time being) because, according to K, 'In my head, it's job done. I got pension, tick : : : It's job done, move on'. This sentiment was echoed by I: 'You just think, tick that box. I am paying into the pension then I am sorted'.

Overriding rational behaviour
The desire for pension scheme enrolment appeared to override rational economic thinking. The employer offered a tax-efficient share saving scheme (SIP) with an employer contribution that was more generous than the pension scheme'sa two-for-one match versus a one-for-one match. A few participants, recognising the fungibility of assets, saw saving into the SIP as a 'no brainer' (D).
If you are in the pension and you are not in the SIP, then you are mad. If you only have enough money to do one, do the SIP because you could actually use that as a pension pot. (O) But SIP participation rates were one-third of pension participation rates. AE likely explains much of this difference, although this also raises questions about the negative influence of AE on asset accumulation noted by Huberman et al.
(). Ditto and Tannenbaum () argue that AE and defaults convey implicit advice. In this case, by automatically enrolling, the employer may have been signalling the desirability of scheme membership and reinforcing the influence of pro-pension norms to the detriment of alternative asset accumulation.

Relevance of the Default Levels
Participants were asked why they thought that the minimum employee contribution level in the pension scheme had been set at %. Several, including H, I, and N, held the view that rates had been set at a low level so as not to 'put people off' (E). They believed that, because scheme membership is voluntary, pension default levels needed to be low to deter the young and the low-paid from opting out. Embedded in these participants' thinking was an assumption that the % default was suitable for younger employees.
I'm thinking of -year-olds or -year-olds. % at that point is probably enough to go on. (A) But there is little evidence that, when making these assumptions, participants knew what level of pension this would provide. Opinions on the suitability of the minimum default were primarily driven by the voluntarism inherent in the UK system. The default was seen as good 'because it gives them something' (F), and was an improvement over not saving at all. This view of the suitability of the default is important because some participants took an active role in encouraging younger employees to remain enrolled (see also Robertson-Rose, ). The appropriateness of the default for younger employees is particularly relevant because of the elasticity of the concept of 'younger'. The participants were all aged over thirty but some still self-identified as 'young'. These young-thinking individuals were more likely to consider current pension arrangements as adequate and to attach little importance to the default parameters. So, for example, there was a contrast between T who was contributing at % and said, I've just turned thirty and I really don't want to be thirty so the last thing I want to do is sit down and talk about pensions. (T) and R who began saving at % because I always thought that me starting a pension at twenty-five, twenty-six years of age I was already five or six years behind the game. (R) The chronological age at which participants viewed it necessary to start saving for retirement and the age at which they believed it necessary to increase contributions varied considerably. This appeared to reflect their perceptions of where they were in the life course, and whether they had dependants or elderly parents.

Antecedent preferences
Strong antecedent preferences are linked to default deviation (Sunstein, ). But this study also suggests that, in addition to inertia and status quo bias, strong current preferences, particularly in relation to children and housing security, are implicated in keeping pension contributions to default levels. Once membership had been achieved, whether additional resources were allocated to the pension scheme depended on where future income security fitted into priorities. Indeed, it could be argued that some were active in accepting the contribution default set by the scheme architecture. They knew it was 'nowhere near enough to be comfortable on a month' (I) but still did not want to increase the amount at their current life stage. Many participants had higher priorities than securing a generous retirement income and were unwilling to restrict experience-related spending, particularly if these experiences might not be available in the future. Although there was evidence that having dependants acts as a driver for both thinking about pension scheme enrolment and contribution increases (because of an unwillingness to burden dependants), having children also curtailed additional saving. A few participants had reduced working hours to spend more time with their children. Irrespective of affordability issues, pensions were a lower priority than the desire not to miss out on child-related opportunities. The conflict between enjoying life now and avoiding future regret at not having saved for retirement was a feature of many interviews. The majority of participants dealt with this internal conflict by mentally allocating future pay rises to their pension. Integral to this approach were assumptions about career development and future pay increases. Thus, behaviour that might be interpreted as myopia and procrastination had its roots in the avoidance of conflicting regrets and in positive illusions about the future.

Insecurity and Alternatives
There was a lack of confidence that, on their own, workplace pensions could provide a satisfactory retirement income. But, rather than significantly increase pension savings beyond the default, there was evidence of intention to diversify investments across different savings vehicles (and asset classes) with the aim of providing multiple sources of retirement income. For example, B, G, I and M contributed the minimum but they also invested in their employer's shares. Although assumptions about life trajectories were generally optimistic, this 'hedging of bets' stemmed primarily from uncertainty about the future. All the participants viewed homeownership as the cornerstone of obtaining security. Indeed, A, U, and W saw property as a more secure asset for retirement than workplace pensions and it was an important component in their retirement strategy.
Aspiring homeowners attached greater priority to saving for a deposit than to increasing pension contributions beyond the minimum. For I, K and M, early mortgage redemption was prioritised over increased pension saving and O had actually reduced pension contributions to fund mortgage debt redemption. Owning one's own house debt-free was viewed as an integral part of managing expenses in retirement. The conversations with the homeowners revealed that there was an emotional response to debt: several, including H and L, expressed pride in the success they were having in becoming mortgage-free. Although participants had similar incomes, housing affordability varied considerably but references to indebtedness were made in absolute terms and not as a percentage of property equity. This would suggest that the symbolism, or the social fact of being debt-free, was of high priority.

Anchoring and Heuristics
Fourteen of the participants were contributing at more than %. Those individuals understood that contributing up to the employer match level meant they were 'maximising their benefits with the company' (C). Concerns about the sustainability of the state pension provision had motivated some to increase contributions but several conveyed the impression that they contributed at the higher level because they felt that it was expected behaviour for individuals of their age and social status. Thus, social norms appeared to be redefining the default rule (Huh et al., ) in participants' minds from the minimum default to the employer's match. But there was also evidence of the match restraining additional savings. Two participants misinterpreted the % level as being the maximum they were entitled to contribute and another assumed that contributions were restricted to multiples of three. The interviews revealed other examples of participants focusing upon the % or % levels. Although most understood that the minimum default would be unlikely to provide a generous pension for higher earners, there was a presumption amongst many that % was a 'correct' figure. Assumptions about the inadequacy of % were based upon the fact that it 'feels a bit low' (E).
Only a couple of individuals mentioned the cumulative employee/employer contribution i.e. % or %, when discussing the amount being credited to their pension scheme. Indeed, in one case, it was only during the interview that the participant recognised that in her previous employment (because of employer generosity) her combined contribution had been four times her current contribution.

I used to pay % into my previous pension, so %, I didn't really question it. (K)
Even when participants recognised that projected fund values might be important, they appeared to be being subconsciously influenced by the scheme architecture and used three and % as reference points to determine the adequacy of their own contribution. There were numerous examples of the figure three being embedded in thinking: participants counting in multiples of three from three to six to twelve. These anchoring effects appeared to be strongest when there was uncertainty about the ideal contribution rate to meet retirement needs.
They can't even show you that if you continue at %, or %, or % that at the end of your time this is the benefit that you will derive from it. (M) Lacking guidelines as to the correct contribution rate, participants appeared to be latching onto the figure of six as an improvement on three. This heuristic is very similar to the interpretation that % is better than zero contributions for younger employees.
A few participants actually pointed to the influence of the default on their contributions.
Whether it's % or %, if it's automatic right from the start, then I wouldn't have known any different. And I would go, right, that's what I have to work with. (G) M and Y both recognised the disconnect between contributing % and obtaining an adequate pension but still appeared to be influenced by the default contribution.

  -
Probably if I had joined and set it at %, it wouldn't have bothered me. But the fact is that it was set at %. It was a % thing to start with and that is what it has been. (M) If the default had been %, I would probably be paying % and I would just have been making do at that level. (Y) Y expressed a frustration that she was losing the benefit of the employer's full matching contribution.
I would prefer it to be at % rather than %. Had they put me at % I wouldn't have changed it down, but having put me at % I felt as if I couldn't justify myself to put it up. (Y) There appeared to be several processes at work reinforcing the status quo bias, including loss aversion (Tversky and Kahneman, ), information failure and decision-making uncertainty. G's willingness to adjust her expenditure to income (which was confirmed elsewhere in the interview) suggests that being made aware of the financial outcome of increasing contributions would have been sufficient incentive to overcome the default. But, for both M and Y, concern about the uncertainty of DC pensions, and the difficulty in determining an appropriate contribution level, reinforced their inertia. The combination of aversion to income reduction (which both expressed) and concern about making the wrong decision, appeared to be reinforcing adherence to a suboptimal default.

Discussion
The first finding that merits further comment is that the study uncovered subtly different motives for adhering to the pension scheme membership and contribution defaults. Enrolment was rarely driven by explicit financial goals. All participants, whether having actively joined the scheme or having been automatically enrolled, expressed normative beliefs about the desirability of membership. Although participants appeared to obtain utility from scheme membership, satisfaction levels were more strongly linked to sociological factors than financial ones. Or in the words of K, 'it was more about just having a pension than the kind of pension.' AE helps realise savings intention but why individuals want to save (or be seen to be saving) in the first place is of considerable relevance. Van Winssen et al. () suggest that anticipated future regret at having made no preparation for negative events is a motivator in health insurance take-up. This study suggests that similar incentives exist with pensions, particularly when the individual considers the risk of future opprobrium. Default adherence is generally investigated from a socio-demographic perspective, being linked to relative youth and low education (Beshears et al., ). However, there is merit in investigating the social and psychological factors that motivate individuals to be complicit in accepting enrolment. One avenue of research may be the subjective meaning attached to membership. Arguably the act of being enrolled (or seeing colleagues enrolled) alters the meaning of the default option so that compliance comes to be seen as a moral act (Davidai et al., ). Psychological identification with a product has been shown to increase the endowment effect (Morewedge and Giblin, ) and it is reasonable to conjecture that the same influences may be at play with pension scheme membership. The author hypothesises that, by saving into the pension, individuals derive social value from being seen to be doing the 'correct thing'. Pension savings have social utility (Beauvois, ) whereas saving into non-pension investment vehicles could be considered individualistic. This may explain the reliance on the pension scheme, rather than the SIP, for asset accumulation. Such behaviour also adds support to the argument advanced by Beshears et al. () that AE can override some individuals' economic rationality. The interrelationship between pension AE and other employer-provisioned savings schemes warrants greater research attention.
Low pension contributions were not always a consequence of inertia and status quo bias. Individuals could be committed to saving for retirement but make active decisions not to commit more resources to the pension. From a long-term income-smoothing perspective, this appears to be sub-optimal, but not if viewed from the perspective of non-financial priorities. It is inherently difficult for an observer to value utility. But the nuanced differentiation between material and experiential consumption suggests that perceptions of experiential opportunity costs are integral to understanding retirement saving decisions. As Loewenstein and Lerner () maintain, emotions can be potent drivers of decision-making. This study suggests that the meaning that we attach to consumption matters. Although participants rarely reflected on the pension income their current savings would provide, it is not clear whether, for some, having that knowledge would have significantly altered their behaviour at that stage in their lives. An individual's present-day preferences do not exist in isolation but coexist with an awareness of their ability to adapt to future circumstances. From that perspective, arguably low contributions are not simply self-regulation failure. Nor is default adherence necessarily sub-optimal in the long run.
The distinction between the intention to save and how much to save may be subtle but is important in the context of low default contribution rates. Individuals' self-assessment of their relative position in the life-cycle appeared to be having some influence on the motives behind their behaviour. The author hypothesises that extended tertiary education and postponed household formation delays self-reflection on the need to save for retirement. Given the importance of early saving for outcomes, the elasticity of the concept of 'younger' warrants further investigation.
Another finding that merits attention is that, once scheme membership had been achieved, the desire for housing security and debt reduction appeared to be of greater emotional and social significance to participants than increasing pension contributions. It is not unreasonable to conjecture that social comparison can influence default adherence. Arguably, it is easier for individuals to demonstrate financial competence by declaring themselves 'mortgage-free' than by increasing pension contributions. Financial literacy shortcomings might explain the preference for mortgage redemption over tax-efficient pension contributions. Yet the participants' expression of pride in their ability to reduce their mortgages suggests that debt aversion is also relevant (Meissner, ). Emotional responses to financial planning are not generally included in financial planning models, although Loewenstein () argues for visceral and situational features to be included in explanations of future-orientated decisionmaking. An alternative hypothesis, drawing on subjective construal theory (McCrea et al., ), is that different levels of abstractness between housing and pensions impact upon proactive behaviour towards contribution defaults. Reducing mortgage debt to zero is an easily understood goal. Achieving pension fund adequacy is much more abstract. The different levels of abstraction could explain the greater willingness to alter mortgage contributions than pension contributions. However, relying on property as a financial safety net in retirement carries risks such as falling house prices, changes to taxation and maintenance costs. A qualitative study investigating individuals' understanding of the relative risks of property and pensions is recommended.
The effects of the default and match rates in this study were illuminating, not only because they restricted savings, as Huberman et al. () have claimed, but also because they influenced discussions about pension adequacy. At one level, some participants were attaching significance to the match heuristicallythey recognised it as an employer benefit. Yet few quantified its value and included it in their pension adequacy calculations. And, at another level, the default and match subliminally inculcated themselves into how participants framed future contributions. This finding is important. It demonstrates the subtle effects of savings adequacy considerations being anchored around the default. Anchoring is particularly prevalent when there is uncertainty (Tversky and Kahneman, ) and the unpredictable outcomes inherent to DC pension schemes could be a clue to the strength of the observed anchoring phenomenon. The dilemma for participants in the study was not that pensions require high financial literacy, but that it was impossible to calculate suitable contribution levels thirty years prior to retirement. Complexity around calculating pension outcomes appears to constrain contributions particularly, as van Schie et al.
() have argued, amongst those who believe they save inadequately. The uncertainty shifted attention to the accumulation of more tangible assets such as housing. Discontent with being in a suboptimal default could not only lead to dissatisfaction with the outcome but also appears to interfere with the accumulation process.

Reflections
This research project used qualitative methods and took a broad exploratory approach to achieve its objective of furthering our understanding of retirement savings behaviour within the context of AE. By focusing on subjective construal, the research has revealed the different motives underpinning pension enrolment and contribution increases. One of the strengths of the paper is that its interdisciplinary approach highlights the importance of combining social, emotional and behavioural factors into pension saving research. The study sheds light on how the combined impact of pro-saving behavioural norms and AE can influence retirement savings behaviour. Given the participants' narrow socioeconomic background, the findings are not generalisable to the wider population. However, UK policy-makers can draw important lessons from the responses of these highly educated, financially secure individuals. It is those on average earnings that have most to lose in the shift from earnings-related to flat-rate state pensions and, consequently, most to gain from increasing workplace pension contributions beyond the default settings. The research draws attention to the danger of expecting individuals to increase contributions spontaneously as they age. An alternative approach could be to introduce automatic age-related increases in contributionsa suggestion supported by the work of Goda and Manchester (), who point to the welfare-enhancing effects of age-related defaults.
The subliminal effects of the default identified in this study are troubling. At a UK-wide level, the minimum default has just recently been raised to %. It will be years before we have large-scale data on the willingness of savers to shift contributions from the default. But it does not auger well that some study participants had been 'stuck' at their employer's default level for over five years. A potential antidote to the stickiness of the default could be to frame workplace pensions as a means to replace a shortfall from the current income position, rather than focusing on contributions. This premise merits investigation, perhaps through a pilot study. On a positive note, the willingness to mentally allocate future earnings to a pension suggests that, were contribution accelerators to become widespread in the UK, these would be received positively. Ultimately, savings vehicles that can tap into individuals' internalised objectives, rather than subconscious behavioural biases, are likely to be more effective and morally more appropriate (Smith et al., ). That said, given that even the financially literate struggle with the complexities of DC pensions, this suggests that there is a need for greater education and guidance on pensions. There is a role here for the Government's new Money and Pensions Service.
To some extent, the reliance on housing as an asset class was an emotional reaction to insecurities about welfare provision in old age. There needs to be a greater understanding that owning one's home outright is seen as integral to providing security in retirement. The shift in the UK to viewing property as a substitute for welfare has received comment elsewhere (Lowe et al., ). Both its causes and consequences in relation to planning and funding in retirement merit greater attention from policy-makers. The preference for using property as a savings vehicle will continue in the absence of state assurances about welfare provision and secure retirement income. Uncertainty is inherent in DC workplace pension schemes and to alter this requires radical changes to current decumulation policy. This is a critical point because, without assurances about reliable retirement income, outcome uncertainty and susceptibility to anchoring will persist. The Government should consider how to inject security into the system, perhaps through encouraging defined ambition pensions where the risk is spread between employer and employee. An alternative approach could be through the provision of Government-backed annuities.
Policy-makers need to ensure that workplace pensions provide adequate and secure retirement income. Automatic enrolment was introduced to the UK based on the success of individual employer schemes in the US. But context is important and UK citizens have different expectations of state support that may limit AE's applicability as a solution to the 'retirement income problem'. The decision to opt out or limit pension contributions to the default does not take place in a cognitive bias vacuum but is influenced by emotional responses to current economic and social policies. A voluntary supplementary pension system set within the context of inadequate housing, precarious employment and derisory elderly care is unlikely to meet the retirement hopes of the majority of the population.