Printer Friendly

Are Kiwis saving enough for retirement? Evidence from SOFIE.

1. Introduction

The extent to which people are saving for retirement is a key element in formulating public policy toward saving and retirement incomes. However, no insight into retirement saving can be gleaned from aggregate measures of household saving rates. It is the accumulation of wealth at the individual level that is the critical indicator. This paper uses new data on assets and liabilities from the Survey of Family Income and Employment (SOFIE) to examine if New Zealand individuals and couples are saving adequately for retirement.

This is a timely question because New Zealand has recently introduced the costly KiwiSaver scheme, whose official motivation is to encourage long-term saving by 'individuals who are not in a position to enjoy standards of living in retirement similar to those in pre-retirement' (Section 3, KiwiSaver Act, 2006). Even within the scheme's first year government expenditure of $1.1 billion was required for KiwiSaver (Treasury, 2008, p. 116) and unexpectedly high expenses on KiwiSaver contributed to the replacement of government surpluses with forecast deficits in late 2008. (1) The KiwiSaver scheme also marked a sharp change from the previous New Zealand policy of tax neutrality for savings (St John, 2007).

The basis for this policy change appears to have been the trend in aggregate measures of household saving rates. For example, the Minister of Finance who introduced KiwiSaver often claimed that 'for every dollar households earn, they spend $1.15 on average' (Cullen, 2007), where this figure appears to come from the difference in 2005 between household disposable income and expenditures in the Household Income and Outlay Account. But as noted above, it is not possible to assess the adequacy of retirement saving from aggregate data.

Such assessment is difficult even with household-level data because few surveys anywhere in the world cover all three topics needed for evaluating the adequacy of saving. Information is needed on assets and liabilities to project wealth through to retirement, on incomes to see what it is that needs to be replaced in retirement, and on either saving or consumption to see what fraction of income is currently being set aside. The survey we use has comprehensive information on assets, liabilities and incomes but only partial information on expenditures so we were forced to use an imputation procedure to estimate consumption and saving. Similar procedures have been used in overseas studies of saving, even in data-rich settings like the United States (e.g. Dynan et al., 2004). Partly because of this caveat of relying on estimated saving rates, and also because of the recent presumption in New Zealand that saving is inadequate, we adopt very conservative choices in our modelling so that any contrary evidence coming from the model should reflect the patterns in the SOFIE data rather than the modelling assumptions.

We develop a formal life cycle model of wealth accumulation which provides estimates of the saving rates that people would need to have until retirement age in order to have an 'adequate' income in retirement. The analysis focuses on ages 45-64 because people in this age range are old enough to start thinking seriously about preparing for retirement. (2) Significant parts of the population aged 45-64 appear to have made adequate provision. This is particularly true among the lower income groups where New Zealand Superannuation (NZS) represents the majority of their retirement income. Apparently, as few as one eighth of the 55-64 age cohort are saving at rates below that needed to accumulate the required level of retirement wealth.

The paper proceeds as follows. The next section briefly describes the data, while an outline of the model follows in Section 3. The findings are set out in Sections 4 and 5. Section 6 concludes.

2. Data

The primary data source in this study is SOFIE, a panel survey which started in October 2002 and is intended to run annually for eight years. SOFIE collects data on levels, sources and changes in income for New Zealand individuals and families. It also reports on major influences on income, such as employment and education experiences, household and family status and changes, demographic factors and health status. The survey covers 26,339 individuals from 10,244 households, representing 3,771,864 people. (3)

The data on assets and liabilities used in this study come from Wave 2, which ran from 1 October 2003 to 30 September 2004. These data contain several limitations, which necessitate assumptions, as outlined below.

* SOFIE's statistical unit is the individual and the household, but the unit of analysis in the retirement model is the non-partnered person and the couple. (4) The couple's income or wealth is made up of the income/wealth of both partners while the age of the couple refers to the age of the older partner.

* Individuals were asked for the total value of each property and the number of other people who also own that property. We assume equal ownership shares among owners.

* There is only one figure that refers to the total value of all mortgages, but no information on the number of mortgages or to which property the mortgages correspond. We assume that the total mortgage value is split between owner-occupied and other residential property such that the loan-to-value ratio is equal between the two classes of property. (5)

* We ignore household items in the calculation of wealth, as these assets depreciate over time and they can not easily be liquidated. These assets are also valued inconsistently across individuals. (6)

* Due to wording errors in the questionnaire, there is evidence that the reported participation rates in pension schemes and values of schemes are markedly lower than indicated by other sources. (7) Since the errors are complex and difficult to remedy, we accept the data as is, acknowledging that these errors understate total net worth by an estimated 2% on average and thus our results would show more undersaving than actually is the case.

3. Saving for retirement--the model

To model adequacy of retirement saving, we adopt a framework of joint determination of saving and replacement rates. This framework assumes that people seek to smooth consumption throughout the life cycle.

3.1. General assumptions

For simplicity, we ignore uncertainty. Specifically, this assumption means that an individual will retire at a certain age as planned; does not engage in the workforce after retirement; knows exactly what their income until retirement will be; can accurately project the rate of return on investments; has a known life expectancy at the age of retirement; knows the amount of NZS that they will receive; plans and executes whatever bequests they wish to make; has no unexpected changes in health status that would affect income or expenditure; and assumes tax rates and other policies remain unchanged. (8)

In the absence of uncertainty, the household chooses a level of consumption that can be financed from income over the working life, and then from savings during retirement. This implies (ignoring interest for the moment) that savings are equal to consumption needs in retirement.

This simple life cycle pattern can be modified to allow for uncertainty. As shown by Moore and Mitchell (1997), when life expectancy is uncertain, consumption will tend to rise until retirement and fall subsequently, rather than remaining uniform throughout. However, the basic pattern of earnings and savings before retirement and wealth decumulation throughout retirement to finance consumption is left unaltered. Moreover, the limited number of studies that explicitly incorporate various forms of uncertainty, such as uncertain earnings realisations, health shocks and end of life uncertainty (Scholz et al., 2006) appear to provide results quite similar to studies that apply financial planning rules of thumb to the same data (Moore & Mitchell, 1997; Love et al., 2009). Therefore, our ignoring uncertainty may not matter too much for the purposes of seeing how adequate, on average, is the retirement provision of a pre-retirement cohort in New Zealand. (9)

Abstracting from uncertainty has the advantage of significantly simplifying the analysis. Clearly, the results cannot be interpreted as applying to a particular individual whose incomes, expenditures, returns on assets and life expectancy are all subject to shocks. However, when these shocks are both unanticipated and distributed equally among both positive and negative changes, the outcomes illustrated here can be interpreted as expected values for any given population group.

3.2. A model of joint determination of saving and replacement rates

The approach adopted follows that of Moore and Mitchell (1997). They argue that it is necessary to develop a model that allows the replacement rate and the pre-retirement saving rate to be jointly determined. The reasons for this are twofold. First, in view of a household's actual and projected income and assets, the saving rate needed to achieve some pre-specified replacement rate may be infeasible. Secondly, the replacement rate depends in part on the rate of taxation in retirement, which in turn depends on the level of retirement income, itself a determinant of the replacement rate. Only when the tax rates in retirement were pre-determined would this second issue be avoided.

The starting point is the condition that real consumption (i.e. income net of taxes and saving) be equal before and after retirement:

[Y.sub.p] - [T.sub.p]- S = [Y.sub.r] - [T.sub.r] (1)


[Y.sub.p] = pre-retirement gross income

[T.sub.p] = pre-retirement taxes

S = savings

[Y.sub.r] = retirement gross income

[T.sub.r] = retirement taxes.

It is argued that under the New Zealand income tax system of TTE, (10) private retirement saving is made from after-tax pre-retirement income [Y.sub.p] - [T.sub.p], and the earnings on the investments are taxed. However, once those accumulated funds are withdrawn (in this case to purchase an annuity), there is no further taxation on the income received in retirement. Furthermore, NZS payments are received net of tax. Hence under this system, [T.sub.r] = 0.

Dividing equation (1) by [Y.sub.p] gives:

1 - ([T.sub.p]/[Y.sub.p]) - (S/[Y.sub.p]) = ([Y.sub.r]/[Y.sub.p]) (2)


R = 1 [t.sub.p] - s (3)


R = [Y.sub.r]/[Y.sub.p] is replacement rate

[t.sub.p] = [T.sub.p]/[Y.sub.p] is pre-retirement proportional tax rate

s = S/[Y.sub.p] is pre-retirement saving rate.


A graphical illustration of the model is presented in Figure 1. At the current time a person/couple has a net worth [W.sub.a] as measured by SOFIE. This wealth is projected to grow to [W.sub.p] by the time they reach a pre-determined retirement age. In order to have a given level of consumption in retirement they would need to have accumulated a stock of wealth equivalent to [W.sub.r]. Part of their retirement income is provided by NZS and the stock of wealth equivalent to the NZS income is incorporated in [W.sub.r] and [W.sub.p].

The difference between the required wealth [W.sub.r] and the projected wealth [W.sub.p] is the shortfall that would need to be accumulated between now and retirement. This additional amount, in the absence of inheritances or unanticipated revaluation in asset values, would need to be built up through savings. These flows are depicted in Figure 1(b).

The approach assumes that some fixed share of pre-retirement income will be saved (s = S/[Y.sub.p]) and the replacement rate is given by the ratio of gross income in retirement to gross income pre-retirement (R = [Y.sub.r]/[Y.sub.p]). Clearly, some values of retirement income could imply a substantial shortfall in retirement wealth, which might in turn require unrealistic or infeasible levels of savings before retirement. It is for this reason that the saving and replacement rates are jointly determined.

3.3. Specific assumptions

We assume that existing policy settings, such as those affecting eligibility for, and the value of, NZS, will be continued. Hence, the retirement age is set at 65 and NZS payments are assumed to grow at 1% annually in real terms, matching the growth rate we assume for average real wages. (11) We estimate age income profiles for each combination of ethnic group/gender/education level and apply the pattern to each individual/couple to project their income until retirement. In estimating the age-income structure, we ignore cohort effects, but allow for an annual real growth rate of 1%, which approximates the average rate of labour productivity growth during the past decade. While higher growth rates are assumed in some models, such as the Treasury's Long Term Fiscal Model, we chose conservative values for all parameters, including the rates of return described below. The income profiles show that income is concave in age, peaking at around ages 45-55 and then declining as retirement age approaches. (12)

Projections of retirement wealth were made on the basis of assumed growth rates defined as after-tax real rates of return for different classes of assets. Net housing wealth was assumed to remain constant in real temps (i.e. no capital gains). Business and financial assets were assumed to grow at 2% per annum, while other classes, such as farms, property and vehicles, were assumed to maintain constant real value. These are conservative values, and allowing for higher rates of return would increase projected wealth and the average share of income that is replaced at retirement, and reduce the proportion of the population without adequate provision. Moreover, we also assume that equity in the primary residence is not consumed during retirement and instead is available as a bequest. This appears to reflect the situation in New Zealand where, according to the Retirement Commissioner: '[E]quity release is a relatively unfamiliar concept for New Zealanders, who are typically reluctant to reduce the capital in their homes after they retire.' (13) This assumption, of not consuming the dwelling, is also consistent with the other conservative assumptions that we have made, since it means that estimated retirement incomes are lower than they would be if housing wealth were tapped into. Finally, whenever discount rates are needed in any calculations or for assets not described above, we assume an after-tax real rate of 2%.

The model for couples is complicated by the fact that the two partners of each couple may neither retire nor die at the same time. The retirement phase for couples is assumed to start when the older partner reaches 65. (The younger partner may continue earning an income, which can affect the value of NZS received by the retired partner.) We further postulate that after one partner dies, the surviving partner will have a consumption level equivalent to 60% of the couple's level.

We compute life expectancies from mortality rates projected by Statistics New Zealand. These projections take into account predicted changes in health status based on "medium' assumptions around fertility, mortality and migration. We assume that Pacific Islanders have the same mortality rates as Maori and that mortality rates for other ethnic groups are the same as for Europeans. As such, we are able to calculate life expectancies at retirement for each gender, broad ethnic group and year of retirement.

4. Saving for retirement--results

The model in Section 3.2 prescribes saving rates as a share of gross income. These figures may not be immediately intuitive: hence, for the empirical results we will report after-tax saving rates. To assess the level of consumption smoothing, we also compute a consumption replacement rate as the ratio of pre-retirement consumption to post-retirement consumption. Some households are prescribed a negative saving rate. Literally, this means that these households should either draw down their current wealth before retirement or borrow against their NZS income to supplement their current consumption, which is hardly feasible in practice. However, those negative saving rates should be interpreted as showing that no further saving is needed to sustain consumption levels in retirement, given the household's current wealth. (14) Even without extra savings, these households would already be able to afford higher consumption in retirement than their present level.

4.1. Baseline results

Table 1 specifies the rate at which households need to save until age 65 so that they could enjoy a level of consumption in retirement similar to what they had before retirement. The required saving rates are both higher and more unevenly distributed among the older households. While the median prescribed saving rate for couples aged 45-54 is 11%, 10% would need to set aside over 35% of their after-tax income for retirement. The 'typical' couple aged 55-64 have a prescribed saving rate of 13%, but at the 90th percentile this rate rises to 44%.

The prescribed saving rates are considerably higher for couples than for non-partnered individuals (Table 2). There are at least three reasons for this. First, the retirement period for couples is longer; it extends from when the older partner retires until when the last partner dies. Second, couples earn more than twice as much as non-partnered people (reflecting the phenomenon of assortative mating), so they have a higher per capita consumption level to sustain. Third, the model does not account for economies of household size in consumption, but NZS does--it pays couples only 54% more than the rate for individuals. (15)

Across the wealth distribution, there is little variation in median prescribed saving rates for the lowest four quintiles. For couples aged 45-54, for example, the median prescribed saving rate ranges from 11% for quintile 1 to 14% for quintile 2, while it is zero for the 20% wealthiest people. For non-partnered individuals, median prescribed saving rates are zero for four out of the five wealth quintiles. These saving rates will enable them to attain a retirement consumption level of around 90% as much as their pre-retirement level. Couples aged 45-54 will expect to have median retirement consumption of $43,100, compared with $34,600 for those nearing retirement.

The prescribed saving rate rises with income level (Figure 2). While the 20% lowest earners should save no more for retirement, the 'typical' household in the top income quintile will need to save around a fifth of their after-tax income to smooth consumption over the life cycle.

The model prescribes no further saving for 34%, of couples and 61% of non-partnered individuals. These households either are earning too little or hold significant wealth. Indeed, 27% of non-partnered individuals and 9% of couples in the sample reported income that was below the current NZS payment; additional saving is not justified for these people as NZS would already provide them more consumption than their present level. Likewise, no more saving is necessary if the household has accumulated sufficient wealth to sustain their pre-retirement consumption levels. In other words, no more saving for retirement is required for those households if they are to retire at 65, given our assumptions in Section 3. They may still need to save for things other than retirement, for a different objective than consumption smoothing, for early retirement, more bequests, or simply as a buffer against uncertainties about health, life expectancy and so on.

4.2. A variation to the baseline

The above results are indeed based on conservative assumptions. First, the level of wealth in private pension schemes reported in SOFIE has been underestimated due to some technical problems with the questionnaire. Second, we assume pre-retirement consumption will be sustained throughout retirement. However, empirical evidence has shown that people consume less in retirement years (Banks et al., 1998, Engen et al., 1999, Brown, 2001). One possible explanation is that increased mortality risk at older ages makes consumption less desirable. (16) Domeij and Johannesson (2006) alternatively hypothesise that the marginal utility of consumption increases with health status. Health depreciates at older ages, lowering the marginal utility of consumption, so consumption spending will necessarily fall. These authors observe that consumption expenditure of Swedish households rises with age until about 60, then declines by 25% by age 80. This finding largely matches the New Zealand pattern documented by Gibson and Scobie (2001).

As a variation to the baseline, retirement consumption is assumed to decline with age. We also impose a cap on retirement consumption; the cap has arbitrarily been chosen to be the 90th percentile value of after-tax income for 64-year-old individuals/ couples. (17) This adjustment avoids treating as inadequate savers those who have saved enough to maintain retirement consumption at the said levels. Under these assumptions, prescribed saving rates are lower. Among the 55-64 age group, over 70% of non-partnered individuals and 50% of couples have no need to make further provision for retirement.

5. Saving adequacy

This section addresses the issue of adequacy. Specifically, we examine the relation between the rate of saving prescribed by the model and an estimate of the rate at which people are actually saving.

5.1. Actual saving rates

In order to conduct the comparison we need data on actual saving behaviour.

Unfortunately, no surveys in New Zealand have been designed to measure savings at individual household level. Remarkably, this lack of data does not prevent many commentators from claiming that New Zealand households do not save 'enough'. Savings can be estimated as income less consumption, (18) but one complication is that SOFIE only collects limited information on expenditures. Therefore, we start by estimating consumption for SOFIE households using the approach suggested by Skinner (1987). Skinner combined demographic and partial expenditure data from the Panel Study of Income Dynamics (PSID) with comprehensive expenditure data from the Consumer Expenditure Survey (CEX) to impute total consumption for PSID households. In particular, Skinner regressed total consumption from the CEX on the consumption elements and demographic variables in the CEX that were also available in the PSID (food at home, food consumed away from home, value of the house, rent, utility payments and the number of automobiles). He then inserted the estimated regression coefficients into PSID data to derive total consumption for PSID households. This method implicitly assumes that a household's total expenditure depends on expenses on food, utilities and rent (for renters) or the house value (for homeowners) and the number of vehicles owned, and that this relationship is constant between the two surveys. This approach and its results have subsequently been applied or extended to derive total household consumption and savings for non-expenditure surveys. (19)

The CEX-equivalent data for our purpose come from the Household Economic Survey (HES). (20) With these data, we are able to adopt Skinner's method to impute expenditure for SOFIE households. (21) We then derive savings as the difference between observed household disposable income and imputed consumption expenditure. Household income by itself is already renowned for having high sampling errors. To make matters worse, SOFIE only collects data on gross income from individuals. We apply the appropriate tax rates to work out disposable income for each person and add up to get household disposable income. That calculation exacerbates the measurement error in the income variable. Consumption expenditure is imputed, so it is also error-ridden. As a result, our estimates of savings are subject to a large margin of error for any individual household although these errors should cancel out when studying the population as a whole or large enough sub-populations. Nevertheless, this is the best that can be done, given the lack of suitable micro data for examining household saving behaviour in New Zealand.



The advantage of this method is that the 'actual' saving rates (as estimated by imputation) relate to the same individuals for whom we have calculated the prescribed saving rates. This contrasts with the method used in Scobie et al. (2005) which could only compare the mean/median actual saving rates from the HES with the prescribed saving rates from Household Savings Survey (HSS) data for broad age groups.

5.2. Comparison

Figure 3 compares prescribed and actual saving rates at seven points along the distribution of saving rates (the 5th, 10th, 25th, 50th, 75th, 90th and 95th percentiles). Overall, actual saving rates well exceed those prescribed by the model. For example, the median prescribed saving rate for non-partnered individuals aged 55-64 is negative, but the median actual saving rate for this group is estimated to be 10%.

Although actual saving rates exceed prescribed saving rates at all quantiles, there may be some people for whom this pattern does not hold since they can occupy different ranks on the distribution of each variable. In fact, under the conservative baseline assumptions, around one third of the population are not saving enough for retirement (see Table 3). Using more realistic adjustments (described in Section 4.2) the proportion of 'problem savers' falls to below 20%. For the group approaching retirement, only 9% of non-partnered individuals and 13% of couples need to increase their saving rates in order to smooth out consumption between now and retirement. This finding is well in line with international evidence. For example, Scholz et al. (2006), who use more sophisticated methods, find that over 80% of Americans are preparing sufficiently for retirement.

6. Conclusions

Both individuals and society have an interest in assuring adequate incomes in retirement. A range of policies including NZS, the Superannuation Fund, the State Sector Retirement Savings Scheme, KiwiSaver, financial education and tax policies are all aimed at addressing the provision of retirement income.

Solid data at the household level must inevitably underpin the foundation of policy. It is widely recognised that income and saving rate measures alone are not an adequate basis for judging final preparedness for retirement. Data on the accumulation of assets and liabilities at the individual level are an essential ingredient of sound policy analysis. After lagging in this area, New Zealand is now developing the data on which more meaningful analysis can be based. The first step was the release of the HSS (2001). It was the first national study of the assets and liabilities of New Zealand households. Subsequently, SOFIE has been initiated. This panel study incorporates a module covering assets and liabilities in every second wave. We use the results of Wave 2 (for 2003/2004) as the basis of the analysis in this paper.

The primary focus is on shedding some light on the question: are New Zealanders saving 'adequately' for retirement? There are two challenging conceptual and measurement issues embodied in this question. The first is how should we define what we regard as 'adequate'? The second is how do we measure the rate at which people are actually saving? Reasonable people may hold a range of views on both matters there is no single 'right' answer.

We have chosen to address the first issue by using a life cycle model where adequacy refers to the ability to maintain one's standard of living in retirement (measured by consumption expenditure) at a level comparable to that enjoyed pre-retirement. For the second issue we have used an indirect method, extrapolating from the HES, to assess the saving rate of those in the SOFIE sample. This step was forced on us, as while SOFIE contains income data, there are no data on consumption and savings per se.

We find that for the majority of people in the lower income brackets no further saving should be required as NZS offers a higher income than their projected pre-retirement income. Likewise, wealthy individuals and couples would not need further accumulation. Overall, 60% of non-partnered individuals and one third of couples are estimated to require no more saving for retirement. After adjusting these baseline results for more 'realistic' assumptions these proportions rise to over 70% of non-partnered individuals and one half of couples.

Somewhere between one eighth and one third of the pre-retirement population have current saving rates below that required for 'adequacy'. Further research is underway to identify the characteristics of this group and to assess the magnitude of their shortfalls, as well as to consider how changes in policy might alter their saving behaviour.


A. The Household Economic Survey (HES)

The HES collects information on household income (both gross and disposable) and expenditure, as well as demographic information on individuals and households. Participants must be New Zealand resident private households living in permanent dwellings. The survey was run annually from 1973 to 1998 and thereafter three-yearly. In this paper we only use data for 2003/2004, to match the timing of SOFIE's Wave 2 data. The 2003/2004 sample contains 2854 households. We made every effort to ensure that the conditioning variables used in equations (4) and (5) are similarly defined between the two data sets.

B. Imputing consumption expenditure and estimating saving rates

Following Skinner (1987), we impute expenditure for SOFIE households by drawing on a similar household survey that has expenditure data, the HES. First, we regress household expenditure on several variables using HES data: (22)

[C.sub.HES] = [[beta].sub.HES] [X.sub.HES] (4)


[C.sub.HES] = household consumption expenditure, as observed in the HES;

[X.sub.HES] = a vector of conditioning variables (disposable income, type, size, number of dependent children, tenure of dwelling and region of residence of the household, and age, education, ethnicity and labour force status of the household head), as observed in the HES. The expenditure elements that are common between HES and SOFIE are dwelling expenses, which include land/water/Body Corporate rates and rent (for renters) or mortgage payments (for homeowners).

We then plug the estimated coefficients [[??].sub.HES] into SOFIE data to predict consumption:


Excluded from our definitions of consumption expenditure are education fees, medical costs, life/health insurance, mortgage principal payments and other capital outlays, as they are investment expenses. There is controversy over how durables (motor vehicles, leisure equipment and household items) are treated. Some argue that durables are stocks that provide flows of services over a number of years, thus treating outlays on durable goods as current consumption will overstate consumption and hence understate savings. Others contend that household expenditures on durables are non-zero households must either own or rent some--so overlooking durables consumption will exaggerate savings.

Ideally, we should account for durables consumption by using a measure of the value of services that the household receives from durables. That measure, termed the annual user cost AUC, can be estimated as:

[AUC.sub.i] - [D.sub.i](r + [[delta].sub.i]) (6)


[D.sub.i] - value of stock of durables type i;

r = risk-free interest rate (after-tax);

[[delta].sub.i] annual depreciation rate for durables type i (available from Inland Revenue Department).

Unfortunately, SOFIE data on [D.sub.i] are unreliable. In particular, values of durable assets in SOFIE were estimated by different methods. (23) Consequently, the stock value of durables is exaggerated, as insured value for replacement and amount that was paid were used over two thirds of the time. Indeed, the average value of household items for people who used amount that would be received if sold is markedly lower than that estimated by insured value for replacement ($6,579 vs. $35,863). While people who used the former may be richer than those using the latter, the gap is too large to warrant that estimation methods have no impact on reported values of durables.

Only 13% used amount that would be received if sold (the 'right' method to use). We selected a sub-sample of households where all members who owned any household items used this method to value those items and estimated their annual user costs of durables using equation (6). Thus, for these household we have three definitions of consumption expenditure:

(a) current consumption expenditure;

(b) current consumption expenditure + outlays on durables;

(c) current consumption expenditure + annual user cost of durables;

and correspondingly three definitions of savings. Saving rates are expressed as a proportion of disposable income.

Of course, the saving rates from (a) are the highest. The average saving rates obtained from (b) and (c) are broadly similar. This is explainable by the fact that in the long run total acquisition costs should be the same as total rental costs. If purchases are evenly distributed across time, then total acquisition costs and total rental costs for each year should also be equal. Since we cannot derive annual user costs of durables for all households, we use saving rates from (b) as baseline estimates of actual saving rates.


We wish to thank Mark Arthur, Lisa Henley, Emma Mawby, Tendayi Nyangoni, Johanna Prebble, Diane Ramsay, Nick Treadgold and John Upfold of Statistics New Zealand for their support with the data. We appreciate the considerable support from the Office of the Retirement Commissioner, especially from David Feslier. Our paper has benefited considerably from the comments of two anonymous referees and the editor. Access to the data used in this study was provided by Statistics New Zealand in a secure environment designed to give effect to the confidentiality provisions of the Statistics Act, 1975. The results in this study and any errors contained therein are those of the authors, not Statistics New Zealand.

(Received 29 January 2008; final version received 15 December 2008)


Attanasio, O.P. (1998). Cohort analysis of saving behavior by U.S. households. Journal of Human Resources, 33(3), 575-609.

Banks, J., Blundell, R., & Tanner, S. (1998). Is there a retirement-savings puzzle? American Economic Review, 88(4), 769-787.

Blundell, R., Pistaferri, L., & Preston, I. (2004a). Consumption inequality and partial insurance. Working papers, Institute for Fiscal Studies.

Blundell, R., Pistaferri, L., & Preston, I. (2004b). Imputing consumption in the PSID using food demand estimates from the CEX. Working papers, Institute for Fiscal Studies.

Brown, J. (2001). Are the elderly really over-annuitized? New evidence on life insurance and bequests. In D. Wise (Ed.), Themes in the Economics of Aging (pp. 91-126). Chicago, IL: The University of Chicago Press.

Charles, K.K., Danziger, S., Li, G., & Schoeni, R.F. (2006). Studying consumption with the Panel Study of Income Dynamics: Comparisons with the consumer expenditure survey and an application to the intergenerational transmission of well-being. Research report, Population Studies Center.

Cullen, M. (2007). Budget gives $40 a week to support saving, Budget speech (May 17).

Deaton, A., & Paxson, C. (2000). Growth and saving among individuals and households. Review of Economics and Statistics, 82(2), 212-225.

Domeij, D., & Johannesson, M. (2006). Consumption and health. Contributions to Macroeconomics, 6(1), 1314-1314.

Dynan, K.E., Skinner, J., & Zeldes, S.P. (2004). Do the rich save more? Journal of Political Economy, 112(2), 397-444.

Engen, E.M., Gale, W.G., & Uccello, C. (1999). The adequacy of household savings. Brookings Papers on Economic Activity, 1999(2), 65-187.

Gibson, J., & Scobie, G. (2001). A cohort analysis of household income, consumption and saving. New Zealand Economic Papers, 35(2), 196-217.

Hubbard, G.R., Skinner, J., & Zeldes, S.P. (1995). Precautionary saving and social insurance. Journal of Political Economy, 103(2), 360-399.

Hubbard, R.G., & Judd, K.L. (1987). Social security and individual welfare: Precautionary saving, borrowing constraints, and the payroll tax. American Economic Review, 77(4), 630-646.

Love, D.A., Smith, P.A., & McNair, L.C. (2009). A new look at the wealth adequacy of older U.S. households. Review of Income and Wealth, 54(4), 616-642.

Moore, J.F., & Mitchell, O.S. (1997). Projected retirement wealth and savings adequacy in the health and retirement study. Working Paper 6240, National Bureau of Economic Research.

Palumbo, M.G. (1999). Uncertain medical expenses and precautionary saving near the end of the life cycle. Review of Economic Studies, 66(2), 395-421.

Paxson, C. (1996). Saving and growth: Evidence from micro data. European Economic Review, 40(2), 255-288.

Scholz, J.K., Seshadri, A., & Khitatrakun, S. (2006). Are Americans saving "optimally' for retirement? Journal of Political Economy, 114(4), 607-643.

Scobie, G.M., Gibson, J.K., & Le, T.V.T. (2005). Household Wealth in New Zealand. Wellington: Institute of Policy Studies.

Skinner, J. (1987). A superior measure of consumption from the Panel Study of Income Dynamics. Economics Letters, 23(2), 213-216.

St John, S. (2007). KiwiSaver and the tax treatment of retirement saving in New Zealand. New Zealand Economic Papers, 41(2), 251-268.

Statistics Canada (2004). Microsimulation models,

The Treasury (2008). Pre-election economic and fiscal update.

Toledo, M. (2006). On the intergenerational persistence of work hours. Meeting papers, Society for Economic Dynamics.

Waldkirch, A., Ng, S., & Cox, D. (2004). Intergenerational linkages in consumption behavior. Journal of Human Resources, 39(2), 355-381.

Ziliak, J.P., & Kniesner, T.J. (2005). The effect of income taxation on consumption and labor supply. Journal of Labor Economics, 23(4), 769-796.


(1.) According to Table 2.5 of the Pre-Election Economic and Fiscal Update, expense changes due to KiwiSaver costs that were not apparent in the May 2008 budget contribute 7% in 2011 and 10% in 2012 of the replacement of government surpluses (of the operating balance before gains and losses) with deficits (Treasury, 2008, p. 30).

(2.) It is more challenging to apply the model to younger ages as the further one is from retirement, the more imprecise projections of retirement wealth, income and consumption become.

(3.) SOFIE's target population is ordinary residents who live in private dwellings. Excluded from the survey sample are short-term overseas visitors (intending to stay for less than 12 months), non-NZ diplomats and diplomatic staff and their dependants, members of non-NZ armed forces stationed in NZ and their dependants, and residents of offshore islands other than Waiheke Island.

(4.) Retirement is an individual's or couple's decision, not a household's. This distinction is also sensible given the structure of NZS payments and their importance to retirement incomes.

(5.) Investment properties usually have a high loan-to-value ratio (for tax benefits), so such division of mortgages would tend to overstate borrowing for owner-occupied properties.

(6.) The methods that were used to evaluate household items include: (1) insured value for replacement (59.4%); (2) insured value not for replacement (6.3%); (3) amount that would be received if sold (13%); (4) amount that was paid (8.1%): (5) other method of estimation (11.7%); (6) don't know; (7) refused: and (8) missing.

(7.) Informal communications and unpublished notes from staff of Statistics New Zealand.

(8.) Uncertainty, including such sources as sickness, disability, employment, earnings, inheritances and life expectancy, can best be introduced using micro-simulation models. See, for example, Statistics Canada (2004).

(9.) For example, Love et al. (2009) use a rule of thumb of having sufficient wealth to generate 150% of poverty-line income over expected future lifetimes and find that only 18% of households in the US Health and Retirement Survey have less wealth than this threshold. Scholz et al. (2006) study a younger cohort from an earlier wave of the same survey, and apply a stochastic life cycle model to calculate the optimal wealth for each household, in the face of various uncertainties, and find that fewer than 20% of households have less wealth than their optimal target.

(10.) TTE refers to a system where the savings are made from after-tax income, the returns are taxed and the withdrawals are exempt. It differs from those systems that exempt savings or earnings from taxation and tax withdrawals (TET, ETT or EET).

(11.) The New Zealand Superannuation Act mandates that NZS payments are kept within a narrow band of net average wages after tax, which is why it is appropriate to assume the same growth rate for future superannuation payments and wages. In terms of the actual growth rate chosen, in keeping with the conservative assumptions used throughout, this growth rate is slightly less than that used in the Treasury (2006) Long-term Fiscal Model, where a growth rate of 1.5% is assumed for average labour productivity and real wages. However, we also note that since year 2000 average labour productivity growth has only been 1.1% per annum (Statistics New Zealand, series: S1LMSP).

(12.) The age at which income peaks and the steepness of the profiles should depend on education, gender, ethnicity, occupation, job status etc, but the data only allow us to account for the first three variables.

(13.) introduction-to-equity-release (accessed 11/12/08).

(14.) We have set negative prescribed saving rates to zero to preclude literal interpretation.

(15.) In 2003, NZS after-tax payment was $12,756 for non-partnered individuals (who live alone) and $19,624 for couples.

(16.) See Banks et al. (1998), Engen et al. (1999), Hubbard and Judd (1987) and Hubbard et al. (1995).

(17.) Equivalent to an annual consumption of $92,000 for couples and $46,000 for non-partnered individuals.

(18.) For examples of estimating saving as the difference between income and consumption see Paxson (1996), Attanasio (1998) and Deaton and Paxson (2000).

(19.) Examples include Palumbo (1999), Blundell et al. (2004a,b), Dynan et al (2004), Waldkirch et al. (2004), Charles et al (2006), Toledo (2006) and Ziliak and Kniesner (2005).

(20.) The survey is briefly described in Appendix A. Some parts of HES annual expenditure are estimated by multiplying by 26 the expenditure information recorded by diary for a household for a two-week period. Therefore, even though expenditure is its primary focus, annual expenditure from the HES is still likely to be measured with errors.

(21.) See Appendix B for further details.

(22.) Equation (4) is estimated in log forms (of income and expenditure). The R-squared is around 60%.

(23.) See Note 6.

Trinh Le (a) *, Grant Scobie (b) and John Gibson (c)

(a) New Zealand Institute of Economic Research; (b) The Treasury, (c) Department of Economics, University of Waikato, New Zealand

* Corresponding author. Email:
Table 1. Prescribed saving rates at various percentiles.


 25th 50th 75th 90th

Non-partnered individuals
Ages 45-54 0 0 10 23
Ages 55-64 0 0 13 34

Ages 45-54 0 11 23 35
Ages 55-64 0 13 31 44

Note: Entries are percentages. Saving rates here are expressed
as a proportion of after-tax income.

Table 2. Median prescribed saving rates, consumption replacement
rates and retirement consumption.

 Ages 45-54

Wealth quintile [] [R.sub.c] [C.sub.r]

1 0 100 12,400
2 0 100 14,600
3 1 99 16,500
4 0 100 18,900
5 0 100 26,800

Overall median 0 100 16,700
1 11 89 32,400
2 14 86 36,700
3 14 86 43,300
4 13 87 52,700
5 0 100 70,400

Overall median 11 89 43,100

 Ages 55-64

Wealth quintile [] [R.sub.c] [C.sub.r]

1 0 100 9,100
2 0 100 11,600
3 0 100 12,400
4 6 94 16,700
5 0 100 23,300

Overall median 0 100 13,600
1 12 88 25,700
2 18 82 30,000
3 19 81 33,300
4 15 85 45,200
5 0 100 67,900

Overall median 13 87 34,600

Note: [] = prescribed after-tax sawing rate; [R.sub.c]
consumption replacement rate; C retirement consumption. Entries
for [S.sub.c] and [R.sub.r] are percentages-.

Table 3. Proportions of the population who appear to be saving
inadequately for retirement.

 Baseline (%) Adjusted (%)

Non-partnered individuals
Ages 45-54 34 18
Ages 55-64 28 9

Ages 45-54 37 26
Ages 55-64 37 13
COPYRIGHT 2009 New Zealand Association of Economists
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2009 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Title Annotation:RESEARCH ARTICLE; Survey of Family Income and Employment
Author:Le, Trinh; Scobie, Grant; Gibson, John
Publication:New Zealand Economic Papers
Geographic Code:8NEWZ
Date:Apr 1, 2009
Previous Article:A report on the 2008 Annual Conference of the New Zealand Association of Economists, Wellington, July 9-11, 2008.
Next Article:Capital intensity and welfare: traded and non-trade sector determinants.

Terms of use | Copyright © 2018 Farlex, Inc. | Feedback | For webmasters