RDP 9209: Financial Liberalisation and Consumption Behaviour Appendix

Data and Replication Results

Data for eight OECD countries, including the United States, Japan, Canada, Australia, the United Kingdom, Germany, France and Italy are collated.

Quarterly total consumption expenditure, disposable income and exports data are obtained from the OECD National Accounts. Total personal consumption less purchases of durables is calculated for the subset of countries (namely, the United States, Japan, France, Italy, the United Kingdom and Canada) which have the appropriate statistics available. All data are measured in per capita terms.

Unemployment, population and interest rate statistics are also obtained from the OECD.

The sample period for each country extends from 1960:1 to 1988:4 with the exception of Japan with a sample period ending in 1988:1.

The unsuccessful attempts to replicate J and P's results are reported only for the two largest countries in their sample, namely, the United States and Japan. Table A1 details our estimates for these countries over both J and P's original sample period and an updated period. J and P source the National Accounts, OECD, Volume II, Detailed Statistics, 1986 for data on Sweden, the United Kingdom, Japan, Italy, Spain and Greece. Equivalent data for the US were taken from The Economic Report of the President (1986).

As mentioned in the text, J and P define consumption to “... exclude expenditure on durables,... [namely] appliances, furniture and means of transportation”, and use the first lag of non-durables consumption, personal disposable income, government expenditure and exports as instruments. However, in order to reproduce J and P's instrumental variables co-efficient estimates for the US, it was necessary to exclude personal consumption expenditure both on durable goods and on nondurable services. For Japan, it was necessary to use national disposable income (which includes the business sector) rather than personal disposable income, and total government outlays as an instrument rather than the national accounts measure of government expenditure (used for the USA). With these inconsistent definitions, it was possible to reproduce J and P's original estimates shown in the last two columns of Table A1. The estimates from consistently defined data are also shown in Table A1, both for the original sample period as used by J and P, and for a larger, more up-to-date sample.

This replication exercise also revealed that the estimates of a₁ are often very close to 1, suggesting that J and P's model (equation (12) in the text) collapses to a linear equation in first differences:

A.l: Tests for Non-Stationarity and Non-Zero Drift

Augmented Dickey-Fuller (1979, 1981) and Phillips-Perron (1988) Z_t tests for non-stationarity are set out in Tables A2 and A3. The ADF procedure also allows a test of the significance of the constant term which indicates the presence of non-zero drift. Notwithstanding the small sample problems associated with these tests,^[12] both the ADF and Phillips-Perron Z_t tests find a unit root in consumption, income and remaining instruments for both the US and Japan at the 5% significance level. The Z_t test reports US Consumption of Non-Durables to be trend stationary at the 10% significance level and the ADF statistic on the constant term cannot reject non-zero drift for this variable alone.

Footnotes

Given the unworkably small size of J and P's samples, the ADF and Phillips-Perron tests are reported over our extended estimation periods. It should be noted, however, that these tests evidence a drop in power when applied to samples of less than 100 observations. [12]