3.4 Stock measure of savings

The 'stock' measure of saving as calculated here comprises changes in real household net wealth as a proportion of real household disposable income.

The total change in net wealth can be viewed as comprising two components: these are sometimes referred to as 'active' and 'passive'. Active saving refers to increases made following a conscious decision by an individual or household, and could take a range of forms including for example term deposits, retirement accounts or investments in shares. The critical issue is that they imply foregone consumption spending.

In contrast, passive saving occurs when there are changes in net wealth arising from the revaluation of assets held by the household. Predominant among these is housing. A rise or fall in house prices will lead to a change in measured net wealth of saving due to real changes in net wealth.

The estimates of stock saving used in this paper are derived from the Reserve Bank's Household financial assets, liabilities and housing values data.[11] In order to estimate the active component of savings it is necessary to remove the effect of asset revaluations from total savings. Ideally this would be done for all assets and liabilities. In reality, data limitations are such that the estimate is restricted to housing, which, in 2011, represented almost 75% of total household assets. The likely consequence of restricting adjustments to asset prices to housing alone is that the 'active' component of saving may be overstated, and the 'passive' component understated by an equivalent amount, however the total would be unaffected.

The following describes the approach adopted. It reflects the fact that housing mortgages may or may not be applied solely to housing but can be used to provide funds for other investments, unincorporated businesses or consumption. Whether the relevant index to indicate purchasing power of mortgage liabilities is the CPI or the HPI depends on the purpose of the mortgage. Two approaches are possible; these are referred to as gross and net.

Measure 1 (gross): V_h/HPI + (A_f - L_h - L_o/CPI (4)

Measure 2 (net): (V_h - L_h)/HPI + (A_f - L_o/CPI (5)

where:

V_h = gross value of housing

L_h = mortgage liabilities

A_f = non-housing (financial) assets

L_o = non-housing liabilities

HPI, CPI = house price index and consumer price index

In the first case, gross housing wealth is deflated by the index of house prices; while in the second net housing wealth is deflated by the index of house prices.[12]The critical difference is that in the first case, mortgage liabilities are deflated by the Consumer Price Index; this case is denoted as “gross”. It is applicable where the house serves as collateral for a mortgage some or all of which can be used for consumption. In the second case the mortgage is viewed as solely tied to the investment in the house, and is denoted as “net”.[13]

Note that the “gross” and “net” adjusted series will tend to diverge if house prices and consumer prices grow at different rates. When the HPI is higher than the CPI, the gross-adjusted series will tend to be below the net-adjusted series. This is because the “real” mortgage will be larger when it is deflated by the CPI (gross adjustment) than when it is deflated by the HPI (net adjustment).

The two measures follow each other closely until the mid 1990's, when they begin to separate as the HPI and the CPI start to diverge with the HPI rising more sharply than the CPI. The purchasing power of saving in terms of the real housing stock has decreased, relative to the real basket of general goods one could purchase.

Each of these measures of housing wealth can be combined with other net assets to form two real net wealth series for the household sector which show the 'active' component of saving. This together with the total real net wealth is depicted in Figure 8.

Figure 8 – Measures of household net wealth: 1978-2011 (constant 1978 dollars)

Source:  RBNZ, QV and authors' estimates

Unsurprisingly, a large proportion of the variation in household net wealth is explained by changes in house prices; this is shown by the passive component of saving. Figure 9 translates these net wealth data into implied household saving rates (using equation (2)). The long run trends in household saving rates using the two measures of housing wealth are shown in Figure 10.

In the most recent two decades, the gross measure of saving has decreased relatively more so than the net measure. This is because house price inflation has tended to exceed CPI inflation, resulting in the real size of mortgage liabilities being larger in the gross calculation.

Figure 9 - Household saving rates based on changes in net wealth (3YMA)

Figure 10 - Household saving after adjusting for changes in house prices (3YMA)

Source:  Source: RBNZ, QV and author's estimates

Asset revaluations (in this case those associated with housing) are a major factor contributing to the difference between flow and stock measures. After excluding the effects of housing revaluations, the stock saving trend appears more similar to the flow measure, albeit with still a less marked downward trend (see Figure 11).

The comparisons between the stock measures based on aggregate data and the flow measures are summarised in Table 5. The results reinforce the finding that stock measures of saving (adjusted for housing revaluations) are substantially higher than the standard flow measure, although they display a similar declining trend.

In their work based on the equity injection method, Hodgetts, Briggs and Smith (2006). estimated implied household saving using an asset and liability acquisitions approach. They concluded that over the period 1987 to 2004, their alternative measure of savings exceeded the HIOA flow measure by an annual average of almost 8 percentage points of household disposable income. In the present study, we find that over the same period the stock measure (adjusted for house prices on a net basis) exceed the average flow measure by 5 percentage points. Both estimates indicate that there is a substantial increase in the estimates of the long run saving rate when a measure based on assets and liabilities is used in contrast to the flow measure.

Figure 11 - Household saving: flow, real stock excluding housing revaluation effects (gross adjustment (A) and net adjustment (B), 3YMA)

Sources: See Appendix F.

a The terms net and gross refer to the system of adjustment for housing price revaluations. See equations (4) and (5).

Delbrück (2012) has updated these estimates and finds that in 2009 the alternative measure of the household saving rates was 8% compared to the household income and outlay measure of -4.1%, or more than 12 percentage points of household disposable income higher.

The results pertaining to the stock measures have, to this point, been based on the aggregate household sector data on assets and liabilities from the Reserve Bank. However as indicated in Table 1, stock measures of saving rates can also be obtained from “micro” data - ie details of assets and liabilities collected from household surveys. The Survey of Family Income and Employment (SoFIE), a longitudinal household survey conducted by Statistics New Zealand, provides extensive coverage of assets and liabilities in every other wave. Estimates of net wealth from this source provide the basis for computing saving rates. Table 6 summaries the results based on SoFIE, the RBNZ aggregate sector data and the flow measures from the Institutional Sector Accounts.

The estimates from SoFIE and the RBNZ, two totally independent sources, are strikingly similar, both for the overall saving rate and the 'active' component which removes the house price revaluations from the overall rate. Further corroboration of these results comes from Le, Gibson and Stillman (2010) who report an estimate 23% for the active component of the saving rate for these years based on SoFIE.[14]

Notes:

1. From Scobie and Henderson (2009), Table 5, p.24. As noted in Scobie and Henderson (2009), the RBNZ data was adjusted to be comparable with the SoFIE estimates. For example the data is converted from December to March years, and the denominator in the saving rate is gross income receivable subtracting imputed rents form the HIOA, as it is gross income which is available in SoFIE. The flow measure here is as a proportion of disposabe income.
2. From Appendix Table C.
3. The overall saving rate is is defined as the change in real net wealth between 2004 and 2006 (in March 2006 prices) divided by the average of 2004 and 2006 real income, divided by 2 to convert to an annual rate.
4. In this comparison of stock measures, business assets (farms, orchards, commercial property) and durables have been excluded from the SoFIE estimates to make then comparable with the RBNZ data which do not include these items.

na: not applicable

A second important conclusion from Table 5 is that again, the stock measure of the household saving rate based in this case on independent survey evidence (SoFIE) is significantly greater than the flow measure from the Household Income and Outlay Account.

A further indication that stock measures provide a different perspective on household saving is summarised in Figure 12. Net household wealth as implied by the flow measure of household saving was estimated, by starting in 1978 with a given stock of net wealth and adding each year the flow of savings. For example if net wealth at the start of the year was say $50bn and savings were $5bn, then the implied net wealth at year end would be $55bn. The results are plotted in Figure 12 together with the actual net wealth series from the Reserve Bank (excluding the effect of house price changes). The flow measure of household saving was apparently negative from 1994 to 2010 (with one exception in 2000).[15]

As a consequence, starting in 1994, the implied stock of net wealth declined every year (bar 2000), to the point that it would now be negative. In other words, the negative saving rates based on the Household Income and Outlay Account imply that households would have completely drawn down their stock of net wealth. However, this is inconsistent with the evidence from both the aggregate sector data on household wealth published by the Reserve Bank, and the micro stock data from SoFIE.[16]

Measures of saving based on net wealth for other similar countries typically yield a “paradoxically” higher and increasing implied stock saving rate than the flow measure, which tends to be declining (see Goh, 2005, and Penner, 2008).

Figure 12 - A stock measure of net wealth compared with net wealth implied by the flow measure of saving