RDP 2006-07: Household Saving and Asset Valuations in Selected Industrialised Countries 1. Introduction

Over the past decade, a fairly synchronised and steady decline in household saving rates has been witnessed in some OECD countries but not in others. A casual inspection reveals that this trend has been most pronounced for four English-speaking industrialised countries – that is, Australia, Canada, the United Kingdom and the United States (Figure 1).

Figure 1: Evolution of Household Saving Rates in Selected OECD Countries
Per cent of disposable income
Figure 1: Evolution of Household Saving Rates in Selected OECD Countries

In this set of countries, the fall in saving since the early 1980s has occurred in conjunction with a trend rise in wealth and improved access to capital gains in both financial and residential housing markets (Figure 2). It appears that gains in wealth have increasingly been used as a substitute for traditional personal saving, that is, saving as measured in the national accounts.

Figure 2: Household Saving Rates and Real Asset Price Dynamics
Figure 2: Household Saving Rates and Real Asset Price Dynamics

To date, the empirical literature on saving has been based on either individual country data (at the macro or micro level) or cross-country panel data. At the country level, most studies have examined developments in the US, where several studies suggest that capital gains – both on equity holdings and on residential real estate – are playing an increasingly important role in the accumulation of household wealth and the fall in household saving. Using aggregate data, Faulkner-MacDonagh (2003) analyses saving ratios relative to wealth in a cointegration framework for the US, and finds that accounting for all sources of wealth (equity, housing, non-equity financial), the US personal saving rate was not far below an implied equilibrium level. At the microdata level, Juster et al (2004) analyse household-level panel data for the US and find that capital gains on corporate equities were the key driver of the decline in the personal saving rate in the US between 1984 and 1994. Likewise, Maki and Palumbo (2001), using microdata, find that the direct wealth effect on household spending can explain much of the decline in the US aggregate household saving rate observed in the 1990s; and that the groups of households that benefited the most from the substantial increases in equity wealth in these years – namely those with high incomes or those who had attained higher education – were also the groups that substantially decreased their rates of saving.[1]

Moving to a multi-country setting, several studies have analysed the macroeconomic determinants of saving in a panel setting. Loayza, Schmidt-Hebbel and Servén (2000) employ dynamic panel data methods with fixed effects estimators for a large panel of countries. They find that there are a variety of factors responsible for underlying trends in saving (though no direct measures of developments in household wealth are included). In particular, using annual data, they find: evidence of inertia in private saving (possibly attributable to lagged effects of the explanatory variables on saving); that the population age structure affects saving through life-cycle channels; and that saving is affected by financial liberalisation, inflation and fiscal policy. A more homogenous group, namely developed economies, is examined in several OECD studies employing panel data techniques which better account for panel heterogeneity, such as de Serres and Pelgrin (2003) and de Mello, Kongsrud and Price (2004).[2] These studies find strong evidence of offsetting movements in private and public saving (or dissaving), lending some support for Ricardian Equivalence. Wealth effects are found to have a common impact on private saving for this group of countries with rising wealth usually associated with lower saving. Sarantis and Stewart (2001) also examine the determinants of private saving in OECD countries, but use panel cointegration tests which allow for both panel and dynamic heterogeneity, although they assume a unique cointegrating vector for the entire panel. They find considerable heterogeneity in the determinants of private saving across OECD countries, ranging from factors such as demographics and credit constraints to public saving. However, they do not include any direct measures of wealth valuation.

While several other studies have analysed issues related to household consumption and wealth effects, relatively few have specifically examined the analogous relationship with household saving.[3] Indeed, analysing household saving is a very complex issue, given both a raft of measurement issues – with household saving as measured in the national accounts being an imperfect measure of life-cycle saving – and a rather complex process through which saving decisions are made.[4] Notwithstanding these issues, a focus on household saving rather than household consumption allows for another perspective on household behaviour, possibly providing some additional insight given that the share of income saved is a metric that many people use to describe their consumption patterns (see Shiller 2004).

This study aims to bridge the gap between existing studies of private saving which have either examined single economies or tended to examine a rather heterogeneous group of countries. It assesses the effect on household saving (as measured in the national accounts) of wealth valuation changes by analysing a subset of countries with similar institutional and cultural features – namely the English-speaking countries identified above – within a common model. The analysis is centered on single-equation error-correction models for each country, with additional estimation for a panel of these countries. The results suggest that there is a significant offset to the decline in saving provided by valuation gains for wealth for these English-speaking countries, that this offset appears to have varied by asset type, and that this offset has become more important through time for some countries.

The rest of the paper is organised as follows. Section 2 discusses some theoretical considerations to justify the basic approach adopted in the empirical analysis. Section 3 details the modelling approach and presents the results. Some concluding remarks are made in Section 4. Several appendices contain background information: Appendix A presents data sources; Appendix B covers statistical measurement issues related to the national accounts measurement of saving; and Appendix C contains background conceptual issues in mapping the national accounts notion of household saving to the accumulation of household wealth.


While a microdata approach may capture effects not present in aggregate data, it can be problematic insofar as findings for a specific cross-section may not be directly applicable to trends at the macro level. That said, a micro approach can provide some complementary information, including household saving behaviour by income category. [1]

The methodologies employed account for panel heterogeneity by allowing the short-run coefficients to vary across countries while imposing the same long-run restrictions. [2]

There is a sizeable literature on wealth effects and household consumption. For example, Bertaut (2002) and Barrell and Davis (2004) employ dynamic cross-country analyses on the basis of single-equation error-correction models, and show a bigger wealth effect on consumption in Australia, Canada, the UK and the US than in other countries, such as those in continental Europe. [3]

Moreover, in the US, periodic revisions have resulted in sizeable changes to historical data on personal saving (see Nakamura and Stark 2005). [4]