report pages 79-84

Section 15 - Households’ financial position – a proper longitudinal survey needed

Sections 9 (tax incentives) and 10 (compulsory private provision) together suggest that governments are relatively powerless to influence directly how and how much citizens save specifically for retirement. Governments can certainly influence, even control, aspects of citizens’ decisions especially where there is compulsory private provision at Tier 2, such as in Australia. They cannot control how households respond to public policy settings in the rest of their financial lives. So, the net effect of those policy settings is likely to be a lot smaller than appears superficially.

We can see what New Zealanders are doing at a ‘macro’ level by analysing total assets and debt available from ‘administrative data’ – information that the government collects for other purposes. The Reserve Bank, for example, produces a report that ‘takes the financial temperature’ of New Zealand’s households as a whole on a regular basis. The most recent report[1]showed households’ gross assets at $1,761 billion[2]with:

  • total liabilities of $218 billion (total debt was 12.4% of gross assets[3]);

  • all housing and land assets (including rental investments) were 47.0% of gross assets;

  • all business investments (unlisted shares and equity in unincorporated businesses) were 24.4% of gross assets[4];

  • KiwiSaver balances ($48.6 billion in 2018) were 5.2% of households’ ‘financial assets’ and 2.8% of all assets.

These ‘administrative data’ do not tell us enough about what happens to individual households and, in the business of pre-retirement planning, that really matters. The only way to discover what is happening at that level is to ask households themselves what they are doing – what they own and owe; what their retirement aspirations might be and what they might be doing about those now and over time. Talking directly to households, rather than individuals, will also show the influence of things that happen to other household members, such as death, divorce, disability or employment-related changes. Those kinds of changes are significant in the retirement and retirement-saving context but are not discoverable from top-down ‘administrative data’.

New Zealand tried to uncover some of this crucial information through the ‘Survey of Family Income and Employment’ (SoFIE). Longitudinal surveys are complex to organise and analyse (and expensive) but can give rich insights into behaviour, particularly changes in behaviour.

SoFIE was a longitudinal survey conducted by Statistics New Zealand over an eight-year period, 2002-2010. It collected financial data about individual New Zealanders every two years during that period, starting in 2004. Because the same individuals supplied information during the whole period, the collection of ‘snapshots’ at each collection date could be ‘joined together’ to give a picture of how participants changed their position over the period[5]. Interestingly, for our purposes, the collection of financial data straddled the introduction of KiwiSaver in 2007.

SoFIE’s sample size started in 2002 at more than 22,000 individuals living in 11,500 households.

We have already summarised some of SoFIE’s insights into New Zealanders’ saving behaviour in the last section 14 (KiwiSaver in the new environment).

SoFIE can also help us to understand how New Zealand’s households as a whole compare with Australia’s. That’s a useful comparison given the very different policy settings in both countries.

The RPRC’s PensionBriefing 2010-2 - What do New Zealanders own and owe? News from SoFIE 2004-2006(accessible here) looked at 2006 data from SoFIE. This was then the most recent information from SoFIE.

The PensionBriefing noted that New Zealanders as a whole had (in 2006):

  • - a lot less in housing of all kinds than many supposed (about 46% of all net assets[6]);

  • - less debt than statistics often represented (about 14% of total gross assets[7]);

  • - more in businesses and financial investments than critics of New Zealanders’ behaviour suggested at the time.

Australia has its equivalent to SoFIE – the Household, Income and Labour Dynamics in Australia (HILDA). HILDA originally (2001) covered 7,682 households and 15,127 adults (age 15+). By 2006, the covered population had reduced to 10,085 individuals. The wealth module was added for the first time in 2002 and became Australia’s first large-scale survey of household wealth since 1915[8].

In 2009, the Australian Government published a commissioned report on data from the then latest wealth module of the HILDA Survey, conducted in 2006. Families, Incomes and Jobs[9]looked at households as a whole in 2006 to see what they owned and owed.

The RPRC’s PensionBriefing 2010-5 (Household Wealth in Australia and New Zealand, accessible here) looked at the Australian equivalent numbers to SoFIE’s. The PensionBriefing used an analysis of the SoFIE 2006 data from a Motu report of “means and medians of assets and liabilities”[10].

Superannuation assets are clearly greater in Australia as a proportion of household’s net assets but business and other assets in New Zealand are greater. There are other differences, but the overall similarities are striking. Given the comparatively similar makeup of populations and institutions, compulsion aside, the overall similarities should probably be expected.

Comparisons of any kind between any two countries are problematic because of the very different environments. For example, with respect to retirement saving and income issues, New Zealand has a universal Tier 1 pension whereas Australia’s is both income- and asset-tested. That affects the way Australians make saving decisions over and above those they are forced to make through the compulsory Tier 2 scheme. If comparisons of retirement saving wealth were to be made, the net value of state Tier 1 entitlements would form an important part on both sides of the Tasman but more so for New Zealanders (relatively more generous pension; no income or asset tests).

Even line-item comparisons of asset ownerships between HILDA and SoFIE are difficult because of different data classifications. With those qualifications uppermost, there seem to be lessons from even a cautious comparison.

Analysing SoFIE itself proved problematic as it progressed through the different waves. Some of the difficulties were noted in the Treasury paper that first reported saving and wealth information from 2006[11]. Also, though both SoFIE and HILDA are longitudinal studies, the RPRC’s 2010 PensionBriefing 2010-5compared just two ‘snapshots’ – the assets and liabilities in both countries for just 2006.

From a retirement saving perspective, what really matters is the net wealth of a retiree at ‘retirement’ and through the retirement period. Aside from the primary residence[12], contents and other ‘lifestyle’ assets (car, boat, etc), the ability to convert other assets to cash (along with state-provided incomes) will drive an individual’s standard of living in retirement[13].

The RPRC’s PensionBriefing 2010-5 analysed the broad split in each country (in 2006) of all the assets that might be available to support all respondents’ retirement income needs, if retirement had occurred in 2006.

In 2006, those totals were:

  • - Australia: 50.5% of total net assets;

  • - New Zealand: 49.4% of total net assets.

The New Zealand number is understated because private superannuation was under-reported (an error in data collection/specification), while family trusts’ holdings, except to the extent there is debt owed by the household to the trust, and Maori assets were both ignored. The split also takes no account of the relative differences in the two state pensions. From a total retirement income perspective, New Zealand retirees need relatively lower amounts of private ‘retirement assets’ than Australians for a given target retirement income.

With those qualifications, as at 2006 (before KiwiSaver started in 2007), the relative similarity of the two overall numbers was notable. Between 1987[14]and 2006, public policy in New Zealand on private provision for retirement was almost completely ‘hands-off’ (no compulsion or tax incentives of any significance). The public policy contrast with Australia could not have been more marked and yet the outcomes, at least as a proportion of the total net assets of all respondents, were relatively similar after about 15 years of consistent public policy (to 2006) in each country.

Australians and New Zealanders had, as a whole, come to relatively similar decisions about the net amounts they had to put aside for retirement, after making allowances for the impact of state interventions in that particular saving project.

Another, more detailed, two-country comparison (Australia and Germany) came to similar overall conclusions. In Living Standards in Retirement: Accepted International Comparisons are Misleading (2011), Melbourne Institute, (accessible here) Joachim Frick and Bruce Headey’s conclusion was that:

“We re-estimate the living standards of retirees in the two countries, following an approach developed by Gruber and Wise... This involves estimating the future lifetime income flows of retirees and integrating these estimates into more conventional ‘stock’ measures of wealth. Also included are estimates of future income-in-kind, notably homeowner imputed rents. The revised ‘present value’ estimates of wealth – ‘comprehensive retirement asset measures’ (CREAM) - suggest that Australian and German retirees are likely to have approximately the same living standards (mean and median), with much the same distribution (Gini).

They observed that “[a]fter many gyrations, our final estimate is that Australian and German retirees have almost exactly the same standard of living [in retirement].”

The differences in public policy settings between Germany and Australia could not be more marked but, again as with the New Zealand comparison, citizens had come to relatively similar decisions overall about the net amounts they had to put aside for retirement after taking account of what the government’s policy settings meant to them personally. That seems a logical outcome and we should actually have been surprised if the outcomes had been materially different.

‘Private’ decisions about what to save for retirement should be ‘net’ decisions; made after allowing for the individual impact of the government’s policies on the individual’s own likely position in retirement. That hypothesis needs testing in New Zealand[15].

Given the potential significance to public policy issues associated with retirement saving and post-retirement welfare, it is surprising there are so few longitudinal studies of households’ financial affairs in other countries. We have already mentioned SoFIE and HILDA. Here is a list of all those we have encountered (and links):

  • Europe: the Survey of Health, Ageing and Retirement in Europe (SHARE) – see here.

  • New Zealand: SoFIE – see here.

  • Australia: HILDA run by the Melbourne Institute– see here.

  • United Kingdom: the English Longitudinal Study of Ageing (ELSA – see here)[16].

  • United States: aspects of households’ financial positions are also covered in the US Health and Retirement Study (HRS)[17]run by the University of Michigan – see here.

That gap in knowledge of what households are doing around the world is notable. What is even more notable is that the Retirement Commissioner’s 2016 Review made no reference to any of the material covered in this section; nor to the importance to public policy considerations of filling our information gaps.

Tax Working Group’s comments

The 2019 Tax Working Group was also constrained by the lack of credible data on household wealth.

“48. The Group considers there is a need for greater public access to data and information about the tax system….It is particularly important to have better data about the distribution of wealth in New Zealand.” (at p. 12)[18].

The TWG recommended, amongst other things, that the government “commission research, using a variety of sources of data on capital income (including administrative data) to estimate the wealth of individuals.” (at page 102).

The longitudinal survey that we contemplate would satisfy the TWG’s recommendation.

Terms of reference for 2019 Review

The Terms of Reference for the Retirement Commissioner’s 2019 Review made no direct reference to the issues raised in this section. However, the Review is asked for information with respect to:

“1. An assessment of the effectiveness of current retirement policies for financially vulnerable and low-income groups, and recommendations for any policies that could improve their retirementoutcomes.

“2. An update and commentary on the developments and emerging trends in retirement income policy since the 2016 review, both within New Zealand and internationally.”

Neither of those “assessments” can be fully addressed without access to the kind of data provided from a full longitudinal study of the kind discussed in this section. We think that if the 2019 Review even raises the issue, that will be a sign of progress.

Questions New Zealand needs to discuss on a new longitudinal survey:

Some of the questions that follow relate to questions raised in earlier sections of this report such as sections 9 (tax subsidies), 13 (occupational superannuation) and 14 (KiwiSaver).

  1. What lessons have we learned from the design and implementation of SoFIE? How can we do it better next time (we think there must be a ‘next time’)?

  2. What lessons might we learn from the longitudinal studies in Australia, Europe, the United Kingdom and the United States?

  3. Do we really know how or how much New Zealanders save for retirement now? Where are they getting advice and what is the extent and quality of that?

  4. Do we really know whether New Zealanders need to save more for retirement than they do now? This will not be answered by asking New Zealanders whether they think they should be saving more but rather examining what they are doing about financial preparation for retirement (including buying and paying off the family home; building a business; acquiring skills that might suit post-‘retirement’ aspirations; potential inheritances etc.).

  5. How do New Zealanders respond over time, both financially and behaviourally, to external changes (global and national financial conditions, labour market changes, technological changes, regulatory changes)?

  6. What impact do family separations and divorce have on New Zealanders’ preparations for retirement?

  7. Do we really know how New Zealanders and their associated households migrate financially from full-time work to full-time retirement?

  8. How does the state pension age of 65 affect New Zealanders’ decision-making about financial preparation for retirement and the retirement decision itself?

There is more on the issues associated with questions 6 and 7 in the next section 16.

[1]Reserve Bank of New Zealand, Household balance sheet, Table C22 as of September 2018 (accessible here).

[2]The Reserve Bank notes that this is an under-estimate of households’ total wealth holdings as it ignores overseas assets and non-life insurance reserves. It also makes no attempt to measure human capital though it does count the student debt ($15.8 billion or 7.3% of all ‘financial liabilities’) that represents part of the cost of acquiring that capital.

[3]The ‘debt to gross asset’ number was 12.4% in 2006, unchanged in real terms over the 12 years to 2018.

[4]These business investments (24.4% of gross assets) were greater than the total assets in bank deposits, securities other than shares, loans, investment fund shares and life insurance/superannuation at 21.9% of gross assets.

[5]The mobility between quintiles of wealth of SoFIE participants was examined in Wealth Disparities in New Zealand – Final Report(2017) by Geoff Rashbrooke, Max Rashbrooke and Wilma Malano, Institute for Governance and Policy Studies – accessible here. In summary, about two-thirds of those in the top and bottom quintiles of wealth stayed there over the eight years measured and about two-thirds of those in the other three quintiles moved (up or down). For such a short period, that seems a relatively high level of mobility. Quintiles of wealth were measured as snapshots in each of 2004, 2006, 2008 and 2010. The groupings are relative to the others at each measurement and do not track overall rises or falls in absolute measures of wealth.

[6]The latest Reserve Bank’s Table C22 as of September 2018 (accessible here)sees that proportion as virtually unchanged over the 12 years 2006 (46%) to 2018 (47%).

[7]The latest Reserve Bank’s Table C22 as of September 2018sees that proportion as having reduced over the 12 years 2006 (14%) to 2018 (12.4%).

[8]Given Australia’s huge intervention in retirement savings with the ‘Superannuation Guarantee’ that was introduced over the 1986-92 period, it seems that Australia had no idea whether Australians needed a compulsory savings scheme. New Zealand did at least have SoFIE data before KiwiSaver started in 2007; it’s just that the government ignored SoFIE’s findings.

[9]Families, Incomes and Jobs: A Statistical Report on Waves 1 to 6 of the HILDA Survey, Roger Wilkins, Diana Warren, Marcus Hahn and Brendan Houng (2009), Melbourne Institute of Applied Economic and Social Research (accessible here).

[10]Household Wealth and Saving in New Zealand: Evidence from the Longitudinal Survey of Family, Income and Employment, Trinh Le, John Gibson and Steven Stillman, Motu Working Paper, Motu Economic and Public Policy Research 10-09 (2010) (accessible here).

[11]Saving Rates of New Zealanders: A Net Wealth Approach, Grant Scobie and Katherine Henderson New Zealand Treasury, (2009) accessible here.

[12]We have made the point above that ‘administrative data’ are not rich enough to capture what is happening at a household level. The ‘primary’ residence is usually a household’s most valuable asset. The most recent Reserve Bank’s Household balance sheet here says that, across all households, all ‘housing and land value’ (excluding, for some reason, vacant land) was 47.0% of gross assets. The equivalent number in 2010 was 49.2%. SoFIE data for 2009/2010 (as reported by Rashbrooke, Rashbrooke and Malano (op cit here, page 17) had respondents’ ‘own home’ as 36.3% of gross assets. The difference between the two numbers (49.2% to 36.3%) is significant and deserves analysis.

[13]Even the primary residence can enter this equation if trading down to a cheaper retirement home is a realistic option. Lifestyle assets could also be sold. Leaving these possibilities aside is a conservative approach.

[14]When tax incentives for retirement saving started their phase-out (completed in 1990).

[15]In Australia, for example, actuaries Milliman Inc have used data to see what pensioners actually spend in retirement. In Big data finally reveals how retirees really live(August 2017, here), retirees aged 65-69 spent $A31,068 in mid-2017. To fund that from all sources, including state benefits, “with 75% certainty” requires capital of about $A130,000 invested in a ‘balanced’ fund. That is a good deal less than many in the financial services industry say is needed.

[16]In Prepared for Retirement? The Adequacy and Distribution of Retirement Resources in England, James Banks, Carl Emmerson, Zoe Oldfield and Gemma Tetlow, Institute for Fiscal Studies (2006 - accessible here), the authors used ELSA data to conclude that most (57%) participants aged between 51 and the state pension age in 2001-02, if they retired immediately, would still have ‘enough’ to live on once they reached state pension age.

[17]The US HRS has a new sample for each survey. While each sample might be representative, it is not a longitudinal survey of the kind we discuss here.

[18]The Future of Tax: Final Report Volume 1 – Recommendations, Tax Working Group, February 2019, accessible here.