This study relates Australian household saving more closely to movements in asset market using event study analysis and econometric analysis. In this study, the policy challenges for Australia from rebalancing in China, a temporary growth slowdown in China, and a recession in advanced countries are analyzed. The Globally Integrated Monetary and Fiscal Model (GIMF) is used for policy challenges. The impact of the mining boom on the Australian labor market is also discussed in this paper.

Abstract

This study relates Australian household saving more closely to movements in asset market using event study analysis and econometric analysis. In this study, the policy challenges for Australia from rebalancing in China, a temporary growth slowdown in China, and a recession in advanced countries are analyzed. The Globally Integrated Monetary and Fiscal Model (GIMF) is used for policy challenges. The impact of the mining boom on the Australian labor market is also discussed in this paper.

I. Why Has Household Saving Increased So Sharply in Australia?1

A. Introduction

1. Australia’s net household saving rate began rising in the mid-2000s and jumped to over 10 percent of gross disposable income after the global financial crisis of 2008–09 (Figure I.1). This was the highest level in nearly 25 years, but was still about 3 percent below the average for the 1970s and early 1980s. The recent spike in saving was higher than in some other advanced economies. Even though the economy is recovering from the crisis, house prices have declined in real terms and the stock market has not fully recovered to its previous highs. In this environment, consumers have remained cautious and households have continued to save and rebuild their balance sheets.

Figure I.1.
Figure I.1.

Consumer Caution in Australia

Citation: IMF Staff Country Reports 2011, 301; 10.5089/9781463921934.002.A001

Sources: Australian Bureau of Statistics (ABS); Reserve Bank of Australia (RBA); Organization for Economic Cooperation and Development (OECD); Haver; and IMF staff calculations.

2. Past staff analysis identified wealth effects, public saving, demography, and the terms of trade as the main factors associated with changes in Australian private sector saving.2 Cross-country regressions on annual data for 19 advanced economies suggest that private saving is negatively correlated with public saving (interpreted as a Ricardian offset) and old age dependency, and is positively correlated with the terms of trade. Single equation estimates for Australia confirm the negative correlation between private and public saving, and show a quantitatively large co-movement between private saving and the terms of trade. They also show a strong negative correlation between changes in household net worth and private saving.

3. This paper relates Australian household saving more closely to movements in asset markets using two approaches:

  • Event study analysis examining deleveraging episodes in a broad sample of advanced economies following different asset market shocks;

  • Econometric analysis using quarterly disaggregated household asset data and staff-constructed asset returns for Australia, Canada, and the United States.

4. The event study analysis suggests that the jump in Australian saving has been large in comparison to past deleveraging episodes in advanced economies. The annual data for 20 advanced economies over 1990–2010 suggests that the pace of deleveraging, and the extent of the jump in the saving rate, is more strongly affected by movements in house prices than in stock prices.

5. Econometric analysis identifies some of the specific channels whereby asset market shocks appear to be related to Australian household saving, and confirms the importance of public saving and the terms of trade. Stock market shocks (which are highly visible) affect saving most strongly in the short term, but housing and especially pension wealth are more important over the medium term. Our error correction specification indicates that a 10 percent decline in Australian house prices from the 2010Q4 level would be associated with a rise in the household saving rate by 1.2 percentage points, while a 10 percent shock to pension wealth would be associated with a similar rise even though pensions are a much smaller portion of wealth. Beyond standard wealth effects, staff-constructed measures of expected medium-term returns to housing are significant, with each percentage point decline in expected returns associated with an increase in the saving rate of ⅓ percentage points. Finally, decomposition of changes in household saving consistently identifies changes in the terms of trade as a quantitatively important contributor.

B. Event Study: Deleveraging Episodes in Advanced Economies

6. For a sample of 20 advanced economies over the period 1990–2010, we identify 12 deleveraging episodes (Table I.1). We use the terminology of deleveraging loosely since actual leverage data is not available for all the countries, and we define such episodes to begin when the household debt to disposable income ratio declines by more than 0.5 percentage points in one year. We restrict attention to episodes for which we have three years of data after deleveraging begins.

Table I.1.

Deleveraging Episodes and Asset Market Shocks

article image

7. The increase in household saving in Australia has been larger than the historical average for deleveraging episodes (Figure I.2). For advanced economies in the past, the gross saving ratio has increased after a negative house price shock, but not after a negative stock market shock alone. The magnitude of the jump in saving in Australia seems unprecedentedly large given that the housing market merely slowed down since the crisis, but did not crash.

Figure I.2.
Figure I.2.

Event Study Analysis

Citation: IMF Staff Country Reports 2011, 301; 10.5089/9781463921934.002.A001

Sources: Australian Bureau of Statistics; Reserve Bank of Australia; OECD; Haver; and IMF staff calculations.

8. Housing market shocks are more likely than stock market shocks to be associated with a deleveraging episode, suggesting that the future behavior of Australian households is sensitive to the evolution of house prices. In the sample, there are eight housing market declines (defined by a 5 percent fall in real house prices), and seven of these were associated with deleveraging episodes. By contrast, only 15 out of the 35 stock market crashes in the data (a crash defined by a 10 percent decline in real stock prices) were associated with deleveraging. Moreover, of these 15 stock market crashes, house prices were concurrently declining by some degree in six of them.

C. Quarterly Saving and Asset Data for Australia, Canada, and the United States

9. To analyze household saving behavior more closely, we collect disaggregated wealth data at a quarterly frequency for a sample of financially deregulated economies. Our analysis covers the period 1990Q1–2010Q4 and we use quarterly data in order to be able to estimate the impact on net household saving of the high frequency movements in different asset prices over this turbulent period. Wealth data for Australia, Canada, and the United States are broken down into housing assets and financial assets (the latter in turn split up into several components including pension wealth and direct stockholdings).

10. Net household saving turned negative in the mid-2000s in Australia, and the rise in the saving ratio both over 2005–08 and since the crisis has been more marked than in the other economies (Figure I.3). Unlike in Australia, net saving rates in Canada and the United States remained positive throughout, reached their nadir (above 1 percent) at a later date and did not show a clear rising trend until the crisis year of 2008.

Figure I.3.
Figure I.3.

Quarterly Data for Australia, Canada, and the United States

Citation: IMF Staff Country Reports 2011, 301; 10.5089/9781463921934.002.A001

Sources: Australian Bureau of Statistics; RBA; Federal Reserve Board; OECD; Haver; and IMF staff calculations.

11. The negative effect of the crisis on household net worth has been only partially reversed, while high saving rates have stabilized both household debt and housing leverage. Household net worth was less affected in Australia than in other countries by the stock market collapse of the early 2000s. The recent crisis sharply reduced net worth, but the housing market rebound in late 2009 helped reverse some of the losses. Pensions and stock market holdings, which are smaller components of household wealth, have not fully recovered. The ratio of housing debt to housing assets spiked with the fall in asset prices but has fallen since then. In the United States, the housing debt to assets ratio remains elevated.

12. Estimates of medium-term housing returns show different patterns for the three economies. Conceptually, households’ saving should be affected by returns not only on deposits but also on the assets they own, such as houses. We assume that households are backward-looking, so that at each point in time, the expected housing return is simply calculated as the trend house price growth over the previous five years. On this measure, Australian expected housing returns reached very high levels by the mid-2000s and have been falling since 2004, well before the global crisis. Both Canadian and the United States expected housing returns were lower during the mid-2000s, and began declining only after the crisis.

13. All three countries suffered a deterioration in the fiscal position of similar magnitude as a result of the global crisis. The stronger fiscal position in Australia and Canada at the onset of the crisis meant that net public saving did not become as negative as in the United States.

14. Australia is unique in the magnitude of the run-up in its terms of trade since the mid-2000s. In part, this is because Australia’s main commodity exports (iron ore and coal) are inputs for the global steel industry, where very strong demand has pushed up prices of raw materials 3–5 times. Canada has also benefited from the global rise in commodity prices, but not to the same extent because oil prices did not rise as much as iron ore and coal prices. The United States, as an importer rather than an exporter of commodities, was hurt by the increase in commodity prices.

D. Econometric Analysis: Panel Regressions

15. The main results of ordinary least squares (OLS) and instrumental variable (IV) regressions, for a panel of quarterly data for Australia, Canada, and the United States for the past 21 years (Table I.2), suggest that:3

Table I.2.

Panel Regressions: Household Saving Rate

article image
Significance: * 10% level, ** 5% level, ***1% level.
  • A 10 percent decline in Australian housing wealth from the 2010Q4 level is associated with an increase in the household saving rate of 1.6 percentage points,4 while a 10 percent decline in pension wealth is associated with a 1.7 percentage point increase (although pensions are a much smaller portion of wealth), according to the lagged OLS specification. This higher sensitivity to pension wealth holds in the IV specification too, and makes sense if households are primarily saving for retirement purposes. Higher sensitivity to financial wealth is consistent with past studies.5

  • A fall in expected housing returns by 1 percentage point is associated with a ¼ percentage point increase in the saving rate. Households may recognize that a lower return to housing (typically their largest single investment) means that they need to save more in order to achieve a given target level of wealth by the retirement date. The slowdown in Australian trend house price growth between 2005 and 2008 may explain part of the increase in saving during this period.

  • Household saving moves in a way that appears to offset about half of any changes in net public saving, according to the lagged OLS specification. This could indicate a Ricardian offset, whereby public borrowing today raises expectations that taxes will increase in the future, inducing households to save. Alternatively, large negative shocks (including, but not necessarily restricted to, the recent crisis) may generate both higher public borrowing and a greater perceived risk of future macroeconomic volatility, and the latter effect may be the driver of increased consumer caution and precautionary saving by households.6

  • A 10 percent increase in the terms of trade is associated with an increase in the household saving rate by 1.5–1.7 percentage points. This robust and quantitatively large impact is surprising and may imply that households view the increase in the terms of trade as temporary, or that the increase in the terms of trade is a harbinger of structural change and makes consumers worried that their jobs (typically outside the mining sector) may be at risk from such change.

  • Demographic variables are not robustly statistically significant in the quarterly specification, perhaps because of higher noise than in the annual data, or because there is less demographic variation in this sample than in past staff work.

16. Structural differences between Australia and the other countries, which may account for saving behavior, are captured to varying degrees by the econometric analysis. It is possible that the saving rate is returning to earlier levels after adjusting to financial deregulation and structurally lower interest rates. To the extent that the effects of deregulation are reflected in movements of assets and liabilities, they should be reflected in the regressions. For the period since the crisis, Australian household incomes have grown faster than those in the United States and Canada, which has increased the scope for deleveraging after a debt-fuelled consumption boom, and may account for the larger jump in the saving rate. The terms of trade variable may be capturing this effect since growth in commodity incomes has been driving the Australian recovery.

E. Econometric Analysis: Single Equation Estimates for Australia

17. The lagged OLS specification for Australia alone yields similar results to the panel regression (Table I.3). The wealth effect from pensions is larger than that from housing: a 10 percent decline in housing wealth from the 2010Q4 level is associated with an increase in the household saving rate of 2.3 percentage points, while a 10 percent decline in pension wealth is associated with a 1.3 percentage point increase. A fall in expected housing returns by 1 percentage point is associated with a ¼ percentage point increase in the saving rate. The terms of trade remains quantitatively important. The graph of residuals for the lagged OLS regression indicates that the model does reasonably well in accounting for the rise in household saving since 2005.

Table I.3.

Australia Regressions: Household Saving Ratio

article image
Significance: * 10% level, ** 5% level, ***1% level.

18. The lower degree of freedom in the single country regression reduces the power of the IV specification. Removing some of the insignificant variables results in an IV regression which qualitatively matches the lagged OLS results, but with larger coefficients.

19. The main factors associated with the increase in the household saving rate since 2005 appear to be the decline in expected housing returns and household asset values, and the rise in the terms of trade. The lagged OLS specification is used for the decomposition and indicates that the fall in expected housing returns was important during 2005–08, while declines in the values of housing and pension assets were associated with a jump in the saving rate from 2008 onward (Figure I.4). The rising terms of trade accounts for quantitatively large increases in the saving rate during the entire period since 2005.

Figure I.4.
Figure I.4.

Lagged OLS Specification for Australia

Citation: IMF Staff Country Reports 2011, 301; 10.5089/9781463921934.002.A001

Sources: IMF staff calculations.

20. An error correction model (ECM) specification for Australian household saving further separates out the impact of different asset prices on saving:

  • In the long term, net household saving is in a cointegrating relationship with housing assets, pension assets, expected housing returns, and the terms of trade (Table I.4).

  • In the short term, stock market shocks (which are highly visible) affect saving, with a 10 percent fall in stock prices being associated with a 0.5 percentage point increase in saving (Table I.5). In addition to the wealth effects from total stockholdings, the specific index of resource sector stocks is important. This short-term relationship quickly gives way to the long-run relationship.

Table I.4.

Long-Run Cointegrating Relationship for the Household Saving Ratio

article image
Significance: * 10% level, ** 5% level, *** level.
Table I.5.

Short-Run ECM Relationship for the Household Saving Ratio

article image
Significance: * 10% level, ** 5% level, ***1% level.

F. Sensitivity Tests and Policy Conclusions

21. Household saving is sensitive to the values of housing and pension assets, and monetary policy can influence consumer behavior by affecting these. Adding real interest rates such as real mortgage rates to the above regressions does not yield statistically significant coefficients, but monetary policy rates are known to affect asset prices and household saving decisions are sensitive to these prices. Moreover, an increase in interest rates would lower net-of-interest disposable income and hence reduce the level of consumption, even if the saving rate remains unchanged. In addition to the wealth effects from housing and pension wealth, expected housing returns are robustly statistically significant. Even if house prices plateau for a prolonged period without falling, expected housing returns would eventually fall and household saving would rise.

22. Some of the Commonwealth government’s fiscal consolidation may be offset by households. Therefore, further policy steps to support household saving may be needed if the government wishes to maintain the current high level of household saving. Adding our crude annual data on superannuation contributions yields mixed results on their effectiveness: member contributions tend to be significant (which means they are not fully offset by households) but employer contributions tend not to be. However, other more detailed studies have identified positive effects from employer superannuation contributions.7

23. According to the decomposition analysis, the terms of trade robustly accounts for a large portion of the changes in household saving, and more research is needed to understand this link. Perhaps households view most of the increase in the terms of trade as temporary (quite different from the government’s view), or the higher income flows may be accruing to individuals in the upper portion of the income distribution who have a high propensity to save. Regarding the post-crisis period specifically, higher income growth in Australia relative to other advanced economies (driven by commodity income) may make it easier for households to begin the deleveraging process, contributing to higher saving. On the other hand, households may view the increase in the terms of trade as a sign that their own jobs may be under threat as the mining sector crowds out their industry of employment, or they may be induced by terms of trade volatility to upgrade their general assessment of economic uncertainty, and therefore to save more. As far as we can tell, the terms of trade is not simply a proxy for the real exchange rate or real oil prices: both are statistically insignificant if added. To the extent that a negative shock to the terms of trade would tend to reduce household saving, consumption would be cushioned from the associated fall in household income. The significant effect of the terms of trade persists even if the sample is restricted to 1990–2004.

References

  • Brittle, S., 2009, “Ricardian Equivalence and the Efficacy of Fiscal Policy in Australia,University of Wollongong Working Paper No. 09-10.

    • Search Google Scholar
    • Export Citation
  • Brooks, R., 2011, “How Well Do Standard Models Explain Australian and New Zealand Saving Behavior?New Zealand Selected Issues Paper, IMF Country Report No. 11/103.

    • Search Google Scholar
    • Export Citation
  • Callen, T., and C. Thimann, 1997, “Empirical Determinants of Household Saving: Evidence from OECD Countries,IMF Working Paper No. 97/181.

    • Search Google Scholar
    • Export Citation
  • Case, K. E., J. M. Quigley and R. J. Shiller, 2005, “Comparing Wealth Effects: The Stock Market Versus the Housing Market,Advances in Macroeconomics, Volume 5, Issue 1, Article 1.

    • Search Google Scholar
    • Export Citation
  • Cashell, B. W., 2005, “Why is the Household Saving Rate So Low?” U.S. Congressional Research Service Series RL33168.

  • Connolly, E., and M. Kohler, 2004, “The Impact of Superannuation on Household Saving,RBA Research Discussion Paper 2004-01.

  • Dvornak, N., and M. Kohler, 2003, “Housing Wealth, Stock Market Wealth and Consumption: A Panel Analysis for Australia,RBA Research Discussion Paper 2003-07.

    • Search Google Scholar
    • Export Citation
  • Favilukis, J., S. C. Ludvigson, and S. Van Nieuwerburgh, 2011, “The Macroeconomic Effects of Housing Wealth, Housing Finance and Limited Risk-Sharing in General Equilibrium,NBER Working Paper No. 15988.

    • Search Google Scholar
    • Export Citation
  • Flavin, M., and T. Yamashita, 2002, “Owner-Occupied Housing and the Composition of the Household Portfolio,The American Economic Review, Volume 92, Issue 1, pp. 345362.

    • Search Google Scholar
    • Export Citation
  • Gong, H., and S. Kelly, 2010, “Overspending or Saving in Property - Do Different Measures of Saving Matter in Australia?University of Canberra NATSEM Online Conference Paper No. CP224.

    • Search Google Scholar
    • Export Citation
  • Klyuev, V., and P. Mills, 2006, “Is Housing Wealth an ‘ATM’? The Relationship Between Household Wealth, Home Equity Withdrawal, and Saving Rates,IMF Working Paper No. 06/162.

    • Search Google Scholar
    • Export Citation
  • Poterba, J. M., R. E. Hall and R. G. Hubbard, 1987, “Tax Policy and Corporate Saving,Brookings Papers on Economic Activity, Volume 1987, No. 2, pp. 455515.

    • Search Google Scholar
    • Export Citation
  • Röhn, O., 2010, “New Evidence on the Private Saving Offset and Ricardian Equivalence,OECD Working Paper No. 762.

  • Tumbarello, P., and S. Wang, 2010, “What Drives House Prices in Australia? A Cross-Country Approach,IMF Working Paper No. 10/291.

1

Prepared by Suman Basu with assistance from Kessia De Leo and Solomon Stavis. The paper benefited from discussions with Ray Brooks and seminar participants at the Australian Treasury Department and the Reserve Bank of Australia.

2

Brooks (2011), “How Well Do Standard Models Explain Australian and New Zealand Saving Behavior?” New Zealand Selected Issues Paper, IMF Country Report No. 11/103.

3

For the OLS specification, one quarter lags are used to reduce endogeneity and standard errors are clustered by country. The IV specification uses lagged values of variables as instruments. The variables in the regressions are non-stationary but the residuals are stationary, implying co-integration in the levels specification.

4

This is calculated by multiplying the 2010Q4 figure for the Australian housing assets/disposable income ratio (which equals 470.2) by 10 percent and then by the regression coefficient in the table of -0.035.

5

See Dvornak, N., and M. Kohler, 2003, “Housing Wealth, Stock Market Wealth and Consumption: A Panel Analysis for Australia,” RBA Research Discussion Paper 2003–07.

6

The coefficient on public saving does not change substantially if the sample is restricted to 1990–2007, so the regression does not capture an effect driven solely by the recent crisis.

7

See Connolly and Kohler (2004), “The Impact of Superannuation on Household Saving,” RBA Research Discussion Paper 2004–01.

Australia: Selected Issues
Author: International Monetary Fund