Patterns of Voluntary Worker Retirement Savings; a Longtitudinal Analysis

JUN FENG, MONASH UNIVERSITY AND PAUL GERRANS, THE UNIVERSITY OF WESTERN AUSTRALIA

12 September 2014

This is the first study using employer-level data to analyse long-term trends in voluntary superannuation contributions in Australia, therefore facilitating the formulation and assessment of superannuation policies. In addition to comparing the role of demographic and social-economic factors in predicting contribution behaviours, we identify the dynamics between salary sacrifice and post-tax contributions, and between past and current year decisions. Interestingly, we identify declining participation in both pre- and post-tax contributions in recent years. We also find that age, income and gender (male) all have a positive association with pre-tax contributions participation whereas for post-tax contributions, income and male gender have a negative effect on average.*

Our analysis provides some contrasting evidence to that previously reported. We find a decline in participation in pre-tax (salary sacrifice) contributions and in post-tax contributions between 2002−03 and 2011−12. The overall participation in pre-tax contributions declined from 24 per cent to 17 per cent, and post-tax contributions fell from 15 per cent to 5 per cent over the period. The 2007−08 financial year is a clear marker with pre-tax contributions recovering before steadily declining after that.

The decline in post-tax contributions is more evident after the 2006−07 financial year (see figure, note Q1 is the lowest income quartile). While there appears to be an association between government policy changes regarding superannuation contributions and pre- and post-tax contributions, there may be other factors at play. The declining trend can also be attributed to lower participation among new cohorts of fund members each year, partly due to changing workforce composition, specifically more part-time workers.

Age, income and gender (male) all have a positive association with pre-tax contributions participation whereas for post-tax contributions, income and male gender have a negative effect on average. The income relationship points to the widening tax incentives for pre-tax contributions as incomes rise, and an attraction to co-contributions for lower income members. A significant interaction effect is revealed between age, gender and income. Female members in the highest income group are more likely to make pre-tax contributions when older.

Conclusions

Private savings for retirement has been an important component in financing consumption after retirement. Following a series of financial market events, policy debates on superannuation and the age pension changes, voluntary savings has become an important factor in ensuring the adequacy of retirement savings. While there has been significant research on the determinants of retirement savings in the US, studies in the Australian context are limited and, where available, they have focused on cross-sectional population surveys, or a single superannuation fund due to data availability.

We examine decade-long contribution records from Australian workers who are members of the Mercer Super Trust database, which provides contribution records from a wide cross-section of employers. The longitudinal nature of the database has provided an opportunity for understanding patterns in voluntary contribution participation over time. The results suggest a positive age pattern (i.e. contributions increase with age) that is consistent with population surveys and behaviour patterns under different institutional settings.

A significant gender differential is observed whereby males are more likely to make salary sacrifice arrangements and less likely to make post-tax contributions. This difference becomes more distinctive when members move to older age groups or higher income groups especially for salary sacrifice arrangements. In both theoretical models and the empirical literature, income is identified as an important factor and is found to be positively correlated with salary sacrifice arrangements. However, surprisingly, it does not hold for post-tax contribution participation.

The contribution records also exhibit a declining trend in participation in voluntary contributions. This pattern is largely due to the substantially lower participation rate among the new members and early leavers in the sub-plans. Separately tracking member cohorts, the results show an increasing trend in voluntary contributions before 2007−08 and a decline afterwards. The longitudinal data also allows us to identify the relation between decisions across types of voluntary contributions and across time. Through regression analysis, the importance of the knowledge of previous contribution behaviours is well established, and suggests the importance of using a longitudinal dataset in the analysis of voluntary contributions. The regression results also indicate that some unobserved characteristics are driving both the decisions of salary sacrifice and post-tax contributions, and that salary sacrifice and post-tax contributions are weak substitutes.

The analysis also provides some new understanding of member behaviours in private retirement savings. Nonetheless, the study also raises a number of puzzles and highlights areas for future work. In particular, the participation pattern identifies a sizably lower participation rate among new members and early leavers. Investigating the underlying reasons for this pattern ─ be it employer specific, market driven, or policy influenced ─ would greatly improve the understanding of such patterns. In addition, innovation in econometric techniques could improve the knowledge of hidden relationships between current and past decisions as well as unidentified traits that drive the contributions decision. This knowledge could provide value to the formulation of superannuation policy. Finally, the analysis has not commented on the amount of contributions. We will address these issues in future work.

* We gratefully acknowledge Mercer Australia for the provision of data for the empirical analysis and specifically David Knox and Dileepa Diyagama in facilitating this and assisting with interpretation of data. Thanks to Jacqui Whale for assistance with data management.