THE IMPACT OF SUPERANNUATION ON HOUSEHOLD SAVING AND WEALTH

Transcription

1 DRAFT THE IMPACT OF SUPERANNUATION ON HOUSEHOLD SAVING AND WEALTH Ellis Connolly and Marion Kohler Preliminary - Please do not quote without permission of the authors June 2003 Economic Research Reserve Bank of Australia The views expressed in this paper are those of the authors and do not necessarily reflect the views of the Reserve Bank of Australia. We would like to thank Luke Gower, Anthony Richards and seminar participants at the Reserve Bank of Australia for helpful suggestions. The authors are solely responsible for the remaining errors.

2 Abstract Over the last twenty years superannuation has grown to be the second largest component of household wealth in Australia after ownership of dwellings. We first examine recent trends in the allocation of household and superannuation fund assets, and we argue that superannuation may help to diversify the household asset portfolio. The substantial rise in contributions to superannuation funds over the 1990s may also have helped to slow the decline in the household saving rate since its peak in the 1970s. In the second part of the paper, after considering other factors influencing household saving, a model is estimated to test whether superannuation contributions have increased household saving. We find evidence that compulsory superannuation contributions have had a positive effect on household saving. JEL Classification Numbers: E21, G11, G2 Keywords: superannuation, household saving, portfolio diversification i

3 Table of Contents 1. Introduction 1 2. The superannuation system in Australia 2 3. Superannuation and household wealth The diversification of the household asset portfolio Factors behind the growth in superannuation assets 8 4. Superannuation and household saving Factors influencing household saving Previous estimates of the superannuation offset Model specification Data and estimation results Data Estimation results Counterfactual saving rate Conclusion 26 Appendix A: Data Definitions and Sources 27 Appendix B: Supplementary results and diagnostics 31 References 34 ii

4 1. Introduction THE IMPACT OF SUPERANNUATION ON HOUSEHOLD SAVING AND WEALTH Over the last 17 years, the Commonwealth government has encouraged Australians to save more for their retirement through a range of pension schemes, including compulsory superannuation and tax incentives for voluntary superannuation. The wide-reaching nature of these policies is likely to have influenced households saving behaviour and the composition of household wealth. Superannuation has grown over the last 20 years to be the second largest component in household wealth in Australia, with a significant effect on the composition of household wealth. In the first part of this paper, we compare the behaviour of households, when they invest directly, to superannuation funds, and draw implications for the diversification of the household asset portfolio. Not surprisingly, the data suggest that superannuation helps to diversify the household asset portfolio in two directions: away from a focus on housing wealth and away from a heavy concentration in domestic assets present in households direct investment decisions. A comparison of superannuation and life insurance assets of Australian households with their United States and United Kingdom counterparts suggests that the growth of superannuation assets is partly attributable to global trends, but also to Australian government policies. Superannuation has not only changed the composition of household wealth, but is also likely to have increased the level of household wealth through flows into households financial assets, despite the continuing decline of the household saving rate in Australia since its peak in the 1970s. In the second part, this paper examines some of the causes of the decline in household saving, and attempts to measure the effect that superannuation growth has had in stemming the slide in the saving rate. Central to this exercise is the offset coefficient, which measures how much of the increase in saving through superannuation has been offset by a decrease in saving through other vehicles. Most estimates of the compulsory superannuation offset in Australia to date have relied on judgment or extrapolation from the experiences of other countries. However, 17 years after the introduction of compulsory superannuation, we have now a sufficiently long time series available to estimate

5 2 the offset coefficient econometrically. We find an offset coefficient of around 38 cents in the dollar, which lies within the range suggested by previous studies. This implies that less than half of the increased saving through superannuation has been offset by a reduction in voluntary saving, thus increasing households saving rate, other things being equal. The next section gives a brief overview of the superannuation system in Australia. Section 3 analyses the impact of superannuation on the composition of household wealth and examines the factors contributing to the growth of superannuation assets. Section 4 analyses the impact of superannuation on household saving, starting with some stylised facts on the determinants of the saving rate of Australian households over the last 40 years, and then estimating a household saving equation with specific focus on the effect of superannuation. Section 5 concludes. 2. The superannuation system in Australia Superannuation remains the dominant long-term policy in Australia for raising household saving, reducing reliance on the age pension (which is provided by the government) and improving retirement incomes. Tax concessions have existed for voluntary superannuation since 1914 (Bateman, Kingston and Piggott (2001), p210). Until the 1980s, interest and capital gains on superannuation funds were not taxed. However, the extent of tax concessions has since been reduced. In 1983, lump sum taxes came into force, while in 1988, a 15 per cent tax on superannuation fund income was introduced, covering employer contributions and fund earnings. In 1997, a further 15 per cent Superannuation Surcharge was imposed on employer contributions for high income earners. In 1986 compulsory superannuation was introduced in Australia. The system initially applied to employees on Federal awards 1 with 3 per cent of their earnings saved in superannuation funds in lieu of wage rises. The system was extended to apply to virtually all employees in 1992 under the Superannuation Guarantee (SGC), with the contribution rate gradually raised to its current level of 9 per cent 1 These are employees whose base wages and conditions were covered by national level arbitration.

6 3 of earnings. These funds will remain preserved until employees retire, which cannot occur until the age of 55 (increasing to 60 by 2025). The scheme has been highly effective in increasing coverage to most employees (Figure 1). 2 Figure 1: The Superannuation Guarantee And Employee Coverage % % 80 Coverage of full-time employees Coverage of part-time employees % % Award superannuation rate Max SGC rate /85 87/88 90/91 93/94 96/97 99/ /03 Source: ABS; ATO Perhaps not surprisingly, households superannuation assets 3 as proportion of GDP almost quadrupled in Australia over the last twenty years (Figure 2), and are now the second largest component of household wealth after non-financial assets. This 2 However, coverage rates for non-employees are considerably lower. The ABS report that in 2000, very few unemployed persons or persons not in the labour force had active superannuation contributions, while almost two-thirds of owner managers of unincorporated enterprises did not have active employer or voluntary superannuation contributions. 3 For simplicity, we use the term superannuation assets for both superannuation assets and life insurance. In line with standard ABS classification, assets in life insurances, part of which is not governed by the superannuation scheme, are reported as part of voluntary superannuation assets. The two items are available separately only since the 1980s, which prevents us from separating them. However, contributions to life insurances that are not superannuation contributions account for about 5 per cent of the total superannuation contributions in the later part of the sample, having steadily fallen from around 15 per cent in the earlier part.

7 4 raises the question what impact has this change had on the household asset portfolio? Figure 2 illustrates another observation: the growth in superannuation funds as per cent of GDP was an experience shared by the US and the UK, which do not have compulsory superannuation schemes in place. Another question is therefore which factors have been driving the unprecedented growth in superannuation assets? We attempt to shed light on both issues in section 3. Figure 2: Household Assets in Superannuation Per cent of GDP % % UK US Australia Sources: ABS; BEA; Federal Reserve; RBA; UK National Statistics. For Australia, the data are households net equity in reserves of superannuation funds and life insurance corporations from the ABS Financial Accounts and Foster (1996) prior to 1988; For the US, reserves in pension funds and life insurance from the Federal Reserve Flow of Funds; For the UK, net equity of households in life assurance and pension funds reserves from the UK National Statistics Financial Statistics and net assets of life insurance and pension funds prior to One of the concerns behind the introduction of the compulsory superannuation scheme was the decline of the household saving rate in Australia since its peak in the 1970s (Figure 3). This raises the issue of whether superannuation has been able to stem the slide in the household saving rate. At first glance, there appears to be a negative relationship since the mid 1980s between household net saving as measured in the National Accounts, and households gross contributions to

8 5 superannuation funds. However, we argue in Section 4 that the fall in household saving may have been exacerbated by measurement problems. Controlling for the roles of capital gains, deregulation and unfunded superannuation, we then estimate a model that allows us to gauge how much of the increase in saving through superannuation has been offset by reductions in other saving components. Figure 3: Household Saving And Superannuation Contributions Per cent of GDP % % 10 8 Household net saving Gross contributions to superannuation funds /60 66/67 73/74 80/81 87/88 94/ /02 Source: ABS 3. Superannuation and household wealth 3.1 The diversification of the household asset portfolio Despite the strong growth in superannuation, the Australian household asset portfolio contains a relatively high proportion of non-financial assets (Figure 4). Ellis and Andrews (2001) examined data over the late 1990s for 10 countries and found that the proportion of non-financial assets in the Australian household balance sheet was close to the highest in the sample. Consistent with this finding,

9 6 in December 2002, almost 70 per cent of Australian household assets were held in housing or consumer durables, compared to around 40 per cent in the US and just over 50 per cent in the UK. Of course, in 2002 the number for Australia is somewhat exacerbated by the fall in share market prices and the large increases in house prices since 2000, but the share of non-financial assets has been consistently above 60 per cent since Figure 4: Household and Superannuation Asset Allocation Households Superannuation funds % % Other Credit market Superannuation Equities Deposits Equities Other Non-financial assets Credit market Non-financial assets Deposits Sources: ABS; RBA. In contrast, 95 per cent of households superannuation was invested in financial assets in December 2002, particularly in bonds and equities (which includes units in trusts, such as property trusts 4 ), asset classes where households have relatively small direct holdings (Figure 4). Increased investment in superannuation may help to diversify the household balance sheet, which in turn can lower the risk of the household wealth portfolio. 4 The standard ABS classification of financial instruments is to include units in trusts in equities.

10 7 Households high allocation of wealth to non-financial assets leaves them more exposed to adverse movements in house prices. The standard mean-variance framework suggests that investing in assets that are uncorrelated with the current portfolio of assets can lower the risk of the portfolio. Table 1 shows annual returns and the correlation of annual returns across the different asset classes from 1980 to Returns on the main classes of financial assets were uncorrelated or negatively correlated with movements in Australian residential property, suggesting that households can reduce the riskiness of their portfolio by investing more in financial assets. While we cannot rule out that the choice of horizon for returns influences this conclusion, it is robust to the use of quarterly or three-year returns. 6 Table 1: Annual Pre-Tax Total Returns on Assets in Household portfolio March December 2002 Residential Treasury Gov t Equities Property US Treasury World equities property notes bonds Trusts bonds Returns Standard deviation Correlation of annual returns Treasury notes 0.00 Gov t bonds Equities Property Trusts US Treasury bonds World equities Sources: Bloomberg, Federal Reserve, MSCI, REIA, RBA, Thompson Financial Datastream. See Appendix A. Investing in superannuation also reduces the weighting of domestic assets in the direct investments of households. Household direct investments are almost entirely domestic, while in contrast, around one fifth of Australian superannuation assets 5 We examine annual returns rather than a shorter horizon since the market for residential property is less liquid than the markets for financial instruments. House price data, especially at a higher frequency, may contain a lot of noise. 6 Nevertheless, even longer horizons may be more appropriate for some assets such as housing and superannuation, over which data become limited.

11 8 are invested overseas. This reduces the exposure of households to Australiaspecific shocks, but at the cost of exposure to exchange rate shocks unless assets held in foreign currency are hedged Factors behind the growth in superannuation assets Households superannuation assets have experienced unprecedented growth over the 1980s and 1990s not only in Australia, but also in the UK and the US. Valuation effects were an important factor in this development, explaining around 70% of the rise in current price terms in the UK and over 60% in the US between 1988 and The importance of market movements is also evident in the reduction of the stock of these assets since 2000 and the more volatile experience of the UK, where equities represented a much larger proportion of assets over the 1990s. However, valuation effects explain only one third of the rise in Australian superannuation assets since 1988, leaving more scope for explanations that flows into these assets have increased. Valuation effects have also influenced the growth in households non-financial assets and direct equity holdings, explaining why superannuation has not grown as significantly as a share of household assets (Figure 4). To abstract from valuation effects, we concentrate on households net flows into superannuation, which reflect movements in contributions, withdrawals and income streams. If households incomes rise, holding the saving rate constant, flows might be expected to rise proportionately. Table 2 therefore shows households flows into superannuation as a share of GDP. We examine the period since 1989 due to data availability, and break it into two equal samples. 8 Australian Households flows into superannuation have grown from an average of 2.8 per cent over , to 4.6 per cent over Notably, over this 7 For details of hedging practices in Australia, see Reserve Bank of Australia (2002). 8 The ABS Financial Accounts began in December 1988, the first comprehensive source of data on household financial assets. From June 1992, the data are from the new Financial Accounts, and prior to June 1992 the data are from the old Financial Accounts. While some classification changes occurred in the new Financial Accounts, these do not appear to have introduced significant breaks in the series we analyse.

12 9 period, US households flows into superannuation fell, 9 while in the UK, households flows remained broadly flat. This suggests that a factor that is specific to Australia has contributed to the rise in flows into superannuation. Table 2: Households' Superannuation Assets Stock of superannuation Per cent of GDP Average net flows into superannuation (a) Per cent of GDP Dec 1988 Dec Australia United States United Kingdom Sources: ABS, BEA, Federal Reserve, RBA, UK National Statistics. Notes: (a) The difference between the change in the stock and the sum of the average net flows over the period reflect not only valuation effects, but also the change in GDP. In Australia during the period, the Superannuation Guarantee was introduced. Subsequently the contribution rate was increased from 5 per cent in 1995 to 9 per cent in In contrast, in the UK and the US voluntary rather than compulsory schemes have been employed to encourage retirement saving. 10 One possible candidate for the difference between countries could therefore be government policies such as compulsory superannuation. The importance of superannuation is also clear from the breakdown of households flows into financial assets into different asset categories (Table 3). Overall, Australian households flows into financial assets have increased from an average of 5.3 per cent of GDP over to 7.4 per cent over Households used superannuation as their primary vehicle for investment in financial assets, with around 60 per cent of household financial flows invested in 9 Part of the fall in flows into US Superannuation may be due to the fact that Individual Retirement Accounts are not included. Household flows into financial assets, which include IRAs, also fell over the period. 10 The main schemes in the US are 401(k) plans, available since 1978, and Individual Retirement Accounts (IRAs), introduced in Both schemes involve tax-deductible contributions, a zero tax rate on income streams until the funds are withdrawn, caps on annual contributions, and restrictions on early withdrawals (see Hubbard and Skinner (1996)). In the UK, contracted out State Earnings-Related Pension Scheme (SERPS) funds have operated since 1978, and were expanded in the 1980s and 1990s. Coverage is voluntary, with around 70 per cent of employees covered. There are minimum contribution requirements. See Bateman, Kingston and Piggott (2001).

13 10 superannuation over the period. Superannuation appears to have contributed to an overall expansion in households financial flows, rather than a mere reallocation from other asset classes, which would imply corresponding outflows from other asset classes. Table 3: Allocation of Households Net Flows into Financial Assets per cent of total financial asset flows Superannuation Currency and Deposits Credit instruments 0-1 Equities (and units in trusts) -1 6 Other 5-3 Sources: ABS, RBA Superannuation has thus grown in importance as an investment vehicle for households. Over the last ten years, superannuation appears to have driven an increase in household financial flows, boosting household wealth and helping to diversify the household balance sheet away from housing. Superannuation policies may have contributed to these developments. However, it is difficult to estimate the effect of superannuation policy using net flows over such a short horizon. Moreover, net flows into financial assets measure only one aspect of households saving behaviour. In the next section, we take a broader perspective and analyse the effect of superannuation contributions on household saving since the 1960s and provide econometric estimates of the impact of superannuation policy on household saving. 4. Superannuation and household saving One of the concerns behind the introduction of the compulsory superannuation scheme was the decline of the household saving rate in Australia since its peak in the 1970s when household net saving reached more than 10 per cent of GDP to around 6 per cent of GDP in the mid-1980s. Compulsory superannuation can increase saving, particularly for two groups of households. One group are liquidity constrained or financially constrained

14 11 households, which consume all or most of their income. For instance, Debelle and Preston (1995) estimate that around per cent of households were liquidity constrained in the 1990s. These households would have had difficulty offsetting compulsory superannuation, unless they were able to borrow. A second group are myopic households who may underestimate how much long-term saving is necessary to accumulate sufficient funds for retirement. A number of surveys have found that households have difficulties in estimating how much saving is needed for an adequate retirement provision. For instance, a recent ANZ Survey of Adult Financial Literacy in Australia (2003) found that only 37 per cent of respondents had worked out how much they needed to save for their retirement. 11 Some evidence of myopic and liquidity-constrained behaviour can be found in the reasons sighted by jobholders for not making voluntary superannuation contributions. Around a quarter of jobholders indicated that they were not interested in making voluntary contributions, while another quarter could not afford to make voluntary contributions. 12 Compulsory superannuation forces these households to save more, unless they are able (and willing) to offset it with either reduced short-term savings or increased borrowing. Compulsory superannuation may be offset by some households, thereby reducing the overall impact on the household saving rate. In principle, if households saving behaviour follows the life-cycle hypothesis, they will respond to forced saving by lowering their marginal propensity to save out of their remaining income in order to maintain their preferred marginal propensity to save out of total income. Therefore, households who save voluntarily could offset the compulsory superannuation contributions (or part thereof) by a reduction in their voluntary savings. Some households may be uncertain about the amount of saving needed for retirement and save voluntarily more than the amount of compulsory superannuation. These households may even reduce their total saving, as they could perceive the amount of compulsory superannuation as a signal of the necessary level. The behavioural responses of these groups, in combination with 11 Also see the ABS Retirement and Retirement Intentions Cat. No , November 1997, which provides evidence that expected retirement incomes are often optimistic. 12 These results are from the Survey of Employment Arrangements and Superannuation in March 2000, reported in ABS Cat no It should be noted that the proportion of respondents not making voluntary contributions to superannuation may have been lower had compulsory superannuation not already existed at the time of the survey.

15 12 the myopic and liquidity-constraint households are factors that influence the overall effect of compulsory superannuation on saving. Voluntary superannuation, without the existence of tax concessions or regulations limiting access to funds, should be close to a perfect substitute for other forms of saving, with few implications for total saving. However, when tax concessions are introduced, voluntary superannuation can provide higher returns than other forms of saving. While tax incentives can encourage households to save more in superannuation, it is less clear whether they increase total saving. Households that would otherwise consume all their income might decide to save in tax-advantaged superannuation to take advantage of the higher returns and maximise lifetime income. However, households that already save voluntarily may merely substitute into superannuation. They may even save less overall since they no longer have to save as much to achieve the same level of lifetime income. After 17 years of compulsory superannuation, and 11 years of the SGC, it may now be possible to look back and analyse the evidence of how the scheme has affected household saving so far. Our main task is to estimate the extent to which compulsory superannuation has been offset by households reducing other forms of saving. We can also examine the offset for voluntary superannuation contributions. 4.1 Factors influencing household saving The household saving rate in Australia has continued to fall since the mid 1980s and therefore, at first glance, there appears to be a negative relationship between household net saving as measured in the National Accounts, and households gross contributions to superannuation funds. However, the fall in household saving may have been exacerbated by measurement problems. Since measurement issues do not explain all of the fall in household saving, we consider the roles of capital gains, deregulation and unfunded superannuation. Some possible control variables are suggested which are then employed in the modelling in Section 3.3. Two sources of measurement problems of household saving have been suggested: inflation bias, and the trend to incorporation (see Edey and Gower (2000), Treasury (1999)). Measured household saving is biased upwards in times of high

16 13 inflation due to the treatment of net interest receipts in the National Accounts. 13 To correct for this problem, household net saving in Figure 5 has been adjusted to reflect the reduced value of households net interest bearing assets (for details, see Appendix A). The adjusted series is much more stable through the 1970s, and has converged to the unadjusted series through the 1990s due to lower inflation and rising household debt. Another potential source of mismeasurement is that the trend to incorporation may have resulted in a downward bias in household saving, as saving by unincorporated enterprises (which is classified as household saving) is gradually being reclassified as saving by corporations. 14 However, there is little evidence of a corresponding rise in the saving rate of non-financial corporations (Figure 5). The mildly inverse correlation between household and non-financial corporations saving over the last 40 years instead may be due to the inflation effect, since households are net interest bearing creditors, while non-financial corporations are net interest bearing debtors. When the inflation-adjusted series are compared since 1989, there is little evidence of an inverse relationship. This suggests that the trend to incorporation may not have been a major cause of the slide in household saving. 13 During times of high inflation, creditors real return on interest-bearing assets, which have a fixed nominal principal, is significantly lower than nominal interest rates would suggest. Since households are net holders of these assets, this leads to an upward bias in measured household saving during times of high inflation, such as the 1970s and 1980s. 14 Other studies that investigate the effect of superannuation on saving, such as Edey and Gower (2000) and Treasury (1999), therefore analyse private saving, rather than household saving. However, the ABS no longer produces private sector saving data.

17 14 Figure 5: Net Saving of Households and Non-Financial Corporations Per cent of GDP % % Households Households inflation adjusted Non-financial corporations Inflation adjusted /60 66/67 73/74 80/81 87/88 94/ /02 Source: ABS; RBA. After making the inflation adjustment, measured household saving still shows some downward trend, although to a smaller extent. Part of this decline may be due to strong capital gains on household assets over the last twenty years, which can be consumed but are not included in the measure of income used in the calculation of household saving. Therefore, actual household saving would be higher if capital gains were included. Since capital gains represent non-income returns to net wealth, we take account of net wealth in our analysis. Financial deregulation has often been cited as a major reason for the slide in household saving, with lower credit constraints allowing households to greatly expand their borrowing. 15 These households have easier access to personal loans or home loans, resulting in lower current disposable income, since households face 15 See, for instance, Edey and Gower (2000) and Ellis and Andrews (2001).

18 15 higher interest payments. 16 Consumption may also increase at least during a transition period since households can reduce their buffer stock saving. And finally some households may aim to increase consumption permanently, by hoping to earn higher returns from their asset portfolio through geared investments, thus increasing lifetime income. 17 All these effects could contribute to a lower rate of household saving. 18 The ratio of household debt to income reflects the increased borrowing possibilities and may therefore be an appropriate measure of those aspects of financial deregulation that impact on household saving (see Bayoumi (1993)). 19 Indeed, this ratio has increased rapidly since the late 1980s, the same period over which the fall in household saving has been most pronounced. Another factor contributing to the fall in household saving is the reduction in reported government superannuation contributions (Figure 6). This is likely to be a result of the gradual phasing out of unfunded government superannuation schemes, for which the ABS imputes contributions into household disposable income. As a result, the total employer superannuation contributions used in calculating household saving have barely increased over the last 20 years despite the rise of compulsory superannuation. In our analysis of the effect of compulsory superannuation on saving we focus on personal saving decisions by households, and therefore remove the ABS measure of employer superannuation contributions from household saving. This has also the advantage that we avoid using a measure that is affected by a range of assumptions which have to be made in order to estimate unfunded superannuation contributions. 16 The cost of servicing a loan of the same size are, however, typically lower as the cost of financial intermediation has decreased due to increased competition and innovation in the financial sector. 17 See Deaton (1992) and Attanasio (1998). 18 These effects at least in part require the borrowing to be financed by other sectors, such as the corporate and overseas sectors. If the borrowing was all financed by the household sector, we would expect only a marginal effect on aggregate household saving. 19 Net wealth may capture this phenomenon only insufficiently, since net wealth is dominated by asset price growth. Moreover, to the extent that debt levels and asset prices are related, these effects could cancel out in net wealth. See Ellis and Andrews (2001) for a discussion of the relationship between deregulation, debt and dwelling prices.

19 16 Figure 6: Employer Contributions to Superannuation per cent of wages and salaries % % 8 6 Total employer superannuation Government superannuation Max SGC rate Award superannuation rate /85 87/88 90/91 93/94 96/97 99/ /03 Sources: ABS; ATO; RBA 4.2 Previous estimates of the superannuation offset The superannuation offset is the extent to which non-superannuation saving falls as a result of increased superannuation saving. Most estimates of the compulsory superannuation offset in Australia to date have relied on judgment or extrapolation from the experiences of other countries, with estimates between 30 and 50 cents per dollar (Table 4). Morling and Subbaraman (1995) estimated empirically the offset coefficient of net superannuation contributions to be 120 cents per dollar or 75 cents per dollar if fund earnings are also included. However, their results are not strictly comparable to the other studies for two reasons. Firstly over the period estimated ( ) superannuation comprised mainly voluntary contributions and secondly their coefficients summarise the behaviours of both the contributing and withdrawing cohorts Contributions to superannuation and withdrawals may not have an exactly inverse effect on household saving since withdrawals could be reinvested, with no net effect on saving.

20 17 In the United States, there has been considerable debate over the effectiveness of voluntary incentives aimed at increasing saving for retirement, with studies using survey data producing conflicting findings. Poterba, Venti and Wise (1996) found large and significant positive effects of tax preferred savings programs on saving behaviour. However, Engen, Gale and Scholz (1996) find little or no saving effects, while Hubbard and Skinner (1996) believe that the truth lies somewhere in between. Table 4: Previous Estimates of the Superannuation Offset in Australia Study Data and Methodology Offset Fitzgerald and Harper (1992) and Fitzgerald (1993) Corcoran and Richardson (1995) Covick and Higgs (1995) Morling and Subbaraman (1995) Gallagher (1997) Examined micro data on number of liquidity constrained households Proportion of employees making sufficient voluntary contributions in 1988 to fully offset compulsory contributions Estimated smoothing of private consumption in 1980s using aggregate consumption function Estimated aggregate relationship between net superannuation flows and other saving over Assumption in RIMGROUP model based on a review of previous studies. 50 cents reduction in net private saving per dollar of compulsory superannuation 17 cents reduction in voluntary superannuation per dollar of compulsory superannuation 37 cents reduction in net private saving per dollar of compulsory superannuation 75 cents reduction in net household saving per dollar of superannuation net flows (120 cents if fund earnings are not included) cents reduction in net private saving per dollar of compulsory superannuation Our study is closest to the methodology chosen by Morling and Subbaraman (1995) in that we estimate the superannuation offset coefficient using annual aggregate data for Australia. However, we estimate both the compulsory and voluntary superannuation offset and we use a different model specification, described in the next section. 4.3 Model specification Figure 7 summarises the influences on household saving as reflected in our estimated model. Voluntary saving is defined as the difference between disposable income and total consumption, where disposable income is the sum of labour and

21 18 investment income excluding capital gains and employer superannuation. Households can choose to save in voluntary superannuation or in other voluntary saving depending on the relative returns. Voluntary saving and compulsory superannuation add to the stock of net wealth. Movements in net wealth, in turn, can reduce voluntary saving through the consumption of capital gains and other withdrawals, such as increased borrowing. The focus of our analysis is on the question what effect do compulsory superannuation and voluntary superannuation have on voluntary saving. Figure 7: Household Saving? Voluntary saving = Disposable income Consumption Other voluntary saving = +? Compulsory superannuation Capital gains and withdrawals Voluntary superannuation Net wealth stock In order to answer this question we formalise this model into a single equation. Given the close relationship between saving and consumption, we have designed our variables to maintain consistency with the work of Tan and Voss (2003) on consumption functions in Australia (See Appendix A for data construction). We define voluntary saving st as a function of autonomous saving β 0, labour income y t, net wealth at the beginning of the period w t and a stationary disturbance term ε t. 21 All variables are after tax, per capita and deflated using the household final consumption expenditure deflator: s t = β 0 + β1yt + β3wt + εt (1) 21 Net wealth at the beginning of the period is used to avoid double counting superannuation contributions, which enter net wealth during the period.

22 19 The introduction of compulsory superannuation csup t forces employers to pay into superannuation funds on top of households labour income. As a result of the increased overall saving, households may reduce their voluntary saving s. A simple way of examining this proposition is to estimate the following equation, and test the coefficient on compulsory saving β, the offset coefficient: 2 t s t = β 0 + β1yt + β 2csup t + β3wt + εt (2) If β 2 is equal to zero this implies no offset, with households unable or unwilling to lower their voluntary saving, and total saving has increased by the amount contributed to compulsory superannuation. If β 2 is equal to minus one, the compulsory superannuation is entirely offset by a reduction in non-superannuation saving, leaving overall saving unaffected. 22 The effect of voluntary superannuation vsup t on other saving can be estimated in the same way, with other voluntary saving os t as the regressand. Finally, as discussed earlier, we capture the effects of deregulation on saving with the debt to income ratio dty t. os t = β + β y + β csup + β w + β dty + β5vsup + ε 0 1 t 2 t 3 t 4 t t t (3) As above, if households consider voluntary superannuation and other voluntary saving as perfect substitutes, we would expect β 5 to be equal to minus one. If voluntary superannuation contributes to (reduces) total voluntary saving at the margin, we would expect β 5 to be larger (smaller) than minus one. The offset coefficient β 2 measures by how many cents non-superannuation saving adjusts in response to an increase of compulsory superannuation contribution by one dollar. One possible scenario would arise if employers superannuation contributions were made in lieu of wage rises. 23 In this case, an alternative 22 Note that in this case, even if overall saving is unaffected, the composition of saving is changed towards saving in long-term assets since compulsory superannuation is typically not accessible before an individual reaches retirement age. 23 It is possible that some compulsory superannuation contributions have increased total labour costs rather than being in lieu of wage rises. Gallagher (1997) assumes that real wages would have only increased by half the amount of the superannuation guarantee contributions if there had been no superannuation guarantee. This would shift the burden of forced saving from households to corporations, potentially having a similar overall effect on national saving.

23 20 benchmark to the offset coefficient would be to estimate by how much total saving has changed compared with the scenario in which the superannuation contribution had been paid out as income. We will call this the net effect of superannuation. For this, reconsider Equation (1). With the introduction of compulsory superannuation, total income becomes y + csup, and total saving becomes s + csup. t t t t s t + c supt = 0 + β 1 yt + c supt ) + β 3 β ( wt + ε t (4) If households behaviour follows the life-cycle hypothesis, they will adjust their marginal propensity to save out of their remaining labour income to maintain the same (overall) propensity to save out of total labour income as existed prior to compulsory superannuation. This is measured as a β1 1 coefficient on compulsory superannuation. Intuitively, this is the amount by which households seek to reduce their voluntary saving in response to compulsory superannuation so that overall, they save of their total income: s β 1 t = β 0 + β1yt + β 1 1) csupt + β3wt + ε t ( (5) If β 2 (from Equation (2)) is equal to β 1 1, this implies that households have adjusted their voluntary saving to offset any excess saving forced by compulsory superannuation. The net saving effect of compulsory superannuation against the counterfactual of a higher income can be measured by subtracting the desired saving rate β 1 from the actual saving rate 1 β 2. Note that the net effect will typically be different from the offset coefficient. The net effect is calculated against the counterfactual that households would have saved the fraction β 1 of every dollar in superannuation if it had been paid as income instead. In contrast, the offset coefficient is calculated against the counterfactual that superannuation contributions would not have been paid to households at all. 4.4 Data and estimation results Data Before presenting the estimation results we discuss briefly the available data on superannuation. Accurate measures of superannuation contributions are difficult to obtain. ATO measured employer superannuation contributions provide us with a

24 21 measure which corresponds to movements in the SGC contribution rate (Figure 8). Data on voluntary contributions (which include contributions to life insurance) are obtained from the National Accounts with an adjustment to remove employer contributions. More detail on the data construction is provided in Appendix A Figure 8: Superannuation and Saving Measures Per Cent of Wages and Salaries % Compulsory % superannuation contributions Voluntary contributions to life offices and superannuation /67 71/72 76/77 81/82 86/87-4 Net inflation adjusted personal saving -8 91/92 96/97 01/02 Sources: ABS; ATO; RBA Estimation results The saving equation is estimated using annual data from 1966/67 to 2001/02. As the data are non-stationary, we estimate the saving equations using Error Correction Models (ECMs). 24 We estimate four models based on equations (2) and (3), both with and without the debt-to-income ratio which proxies for financial 24 ADF tests suggest all the variables are I(1). The speed-of-adjustment coefficients in the ECMs are negative and significant, consistent with the presence of cointegration. The speed-ofadjustment terms of between 1 and 2 suggest that voluntary saving overshoots within one year in response to a shock. Overshooting of saving with respect to income, for instance, is consistent with a model where consumption does not adjust instantaneously to permanent shocks to income.

25 22 deregulation. The results for the long-run coefficients are presented in Table 5, with more detailed results, including further diagnostics, reported in Appendix B. Table 5: ECM Results For Saving Equations: Long Run Coefficients Sample: 1966/67 To 2001/02 Voluntary saving Voluntary saving ex super Labour income 0.08* (0.03) Compulsory superannuation ( β 2) Voluntary superannuation (a) (b) (c) (d) -0.58* (0.32) Net Wealth -0.01* (0.01) Debt to income in per cent Cointegration -1.10** (0.28) 0.13** (0.03) (0.26) 0.00 (0.01) -0.02** (0.01) -1.29** (0.20) 0.09* (0.04) (0.32) -1.30** (0.23) (0.01) -1.31** (0.26) 0.13** (0.03) (0.21) -1.30** (0.19) 0.00 (0.01) -0.02* (0.01) -1.47** (0.24) 2 R Chow breakpoint {0.44} {0.60} {0.47} {0.47} (1988/89) Wald tests on β 2 : no offset ( β 2 = 0) {0.07} {0.15} {0.15} {0.13} total offset( β 2 = 1) {0.19} {0.02} {0.10} {0.00} Notes: Numbers in parentheses () are standard errors and **, * represent significance at 5 and 10 per cent levels. Standard errors on the long run variables in the ECMs are calculated using a Bewley Transformation and are Newey-West corrected for heteroskedasticity. Numbers in braces {} are p-values. Cointegration is a test on the speed-of-adjustment term. In regression (a), net inflation-adjusted voluntary saving is modelled as a function of labour income, compulsory superannuation and net wealth. The marginal propensity to save out of labour income is around 8 cents in the dollar. When the debt-to-income ratio is also included as a regressor in regression (b), the marginal propensity to save out of labour income rises to around 13 cents in the dollar. These estimates are somewhat lower than the marginal propensity to save which can be implied from the consumption function estimated by Tan and Voss (2003), of around 30 cents in the dollar. The difference may be due to Tan and Voss

26 23 estimation period, which starts only in the 1980s, or the treatment of consumer durables, which are excluded from Tan and Voss measure of consumption, and from our measure of saving. 25 The point estimate of the offset of compulsory superannuation in regression (a) is somewhat high, at 58 cents per dollar. However, compulsory superannuation was introduced around the same time as financial deregulation began. Consequently, we could have an omitted variable bias on the offset coefficient if we do not control for the effects of financial deregulation on saving and borrowing. 26 Regression (b) confirms this suspicion: the debt-to-income ratio is highly significant, and the offset coefficient falls to 38 cents per dollar. This estimate is within Gallagher s expected range of 30 cents to 50 cents. A Wald test suggests that this offset is significantly below a full offset of minus one and not significantly different from no offset at all. Regressions (c) and (d) are the equivalent to equation (3), which explains other voluntary saving with the same variables as in (a) and (b), plus voluntary superannuation. The offset coefficient on voluntary superannuation is quite large at 130 cents, but close to the 120 cents estimated by Morling and Subbaraman (1995). Since Wald tests are unable to reject that this coefficient is equal to 1, these estimates suggest that contributions to voluntary superannuation have roughly been offset by decreases in other voluntary saving. However, these results are likely to be affected by the quality of the data, with some double counting of voluntary superannuation contributions through rollovers likely to bias the size of the voluntary superannuation offset upwards (more details in Appendix A). We should also note that our measure of voluntary superannuation includes life insurance and has been subject to a number of changes in the tax regime over the sample period. 25 The coefficient on net wealth in (a) is also smaller than the implied coefficient in Tan and Voss. However, the two coefficients are not strictly speaking comparable. In their consumption framework, net wealth increases consumption through both capital gains and interest income, while in our saving measure net wealth would be expected to have a negative effect on measured saving through capital gains, but a positive effect through higher interest and dividends. 26 Several other variables were introduced into the model to test for potential omitted variables bias. These variables, which included the real interest rate, demographic variables and measures to capture uncertainty, such as the unemployment rate, were found to have an insignificant effect in our saving equations. More details can be found in Appendix B

27 24 The other variables do not appear to be significantly affected by the introduction of voluntary superannuation as a regressor in (c) and (d). The coefficients on compulsory superannuation in regressions (c) and (d) are slightly lower than the corresponding coefficients in (a) and (b). The lower coefficients may suggest that households reduce their voluntary superannuation contributions to offset compulsory superannuation, but we should note that the wide standard errors make this only a tentative conclusion. Some supporting evidence that compulsory superannuation may have an offsetting effect on voluntary superannuation can be gleaned from the Household Expenditure Survey. Although the total value of voluntary contributions has grown since the 1980s, this has been driven by highincome earners, while lower to middle income earners reduced their contributions. Households who may have otherwise chosen to save in voluntary super may be increasingly relying on their compulsory contributions Counterfactual saving rate In this section we construct a counterfactual to gauge the net effect compulsory superannuation has had on the saving rate. Of course, such a scenario analysis can always be only illustrative as we have to hold all other factors constant. We consider the alternative scenario that compulsory superannuation had been paid out as income to households (Equations (4) and (5)). Based on the estimated equation (b) for total voluntary saving, an increase in compulsory superannuation of one dollar leads to an overall increase in saving of 62 cents. On the other hand for every dollar in additional income households would have saved 13 cents voluntarily. The net effect of one dollar in compulsory superannuation is therefore estimated to be 49 cents. For completeness, the net effect on the basis of equation (a) is 34 cents. If we assume that these marginal effects are also the average effects, we can construct a rough counterfactual household saving rate, which is depicted in Figure 9. A 49 cent in the dollar contribution appears to have contributed slightly over 1 per cent of GDP to the saving rate by 2001/02. This is roughly in line with the model calibrations by the RIM Task Force, which estimated that the effect of compulsory superannuation on the national saving rate would be an increase of around 1.5% by 2002 (see Edey and Simon (1996), Figure 7).