Wealth Effects and Consumption: A Panel VAR Approach

Abstract

We provide new evidence on the comparison between the stock and housing wealth effects on consumption. Using a panel VAR approach applied to OECD data, we find evidence that the stock market wealth effect is generally the larger. However, with regard to the evolution of asset wealth effects over time, our findings show that the housing wealth effect has outweighed the share market wealth effect in the last decade. We further find that asset wealth has asymmetric effects on consumption, with stronger and more persistent effects from positive asset wealth shocks. Our results have important monetary policy implications for both stock and real estate markets, and offer timely insights into the desirability of current proposals to reduce house price volatility, such as through macro prudential regulations.

Keywords:

• consumption
• wealth effect
• stock price
• housing price
• asymmetry
• panel vector autoregression

JEL Classifications:

• C33
• E21
• E44
• G15

Acknowledgements

We are most grateful for helpful suggestions provided by two anonymous referees as well as participants at the 2013 meeting of the Australian Conference of Economists. Any remaining errors are our own.

Notes

1. We also considered alternative orderings. For example, placing asset wealth in front of consumption, did not lead to any qualitatively different results from the Panel VAR estimation.

2. A panel VAR is estimated here instead of a panel VECM because earlier panel cointegration test results showed that only weak evidence of a cointegrating relationship between consumption, income, financial and housing wealth is presented. Results are available upon request.

3. The main focus of this paper is on the impulse response of consumption resulting from shocks to asset wealth. Given that an optimal lag length of four is used in the GMM estimation of both the four- and six-variable VAR models, the full results for equation (1(1) ) are not reported here in order to conserve space. These estimates are available from the authors upon request.

4. We set the break at the year 2000 for two reasons. First, pretesting based on the Bai and Perron (2003) structural break test applied to each country showed that the break date across countries ranges from 1998 to 2002 if only one break date was allowed (test results are available from the author upon request). Therefore, we choose the mid-point of the year 2000 as the break date. Second, the housing market boom started in the early 2000s and the stock markets collapsed at the same time due to the dot-com bubble bursting in 2001. Moreover, this period was also characterized by the start of the common monetary policy in 1999 and the introduction of the euro in 2002.