On testing Easterlin's hypothesis using relative cohort size as a proxy for relative income

✍ Scribed by Tilak Abeysinghe

• Publisher: Springer
• Year: 1991
• Tongue: English
• Weight: 933 KB
• Volume: 4
• Category: Article
• ISSN: 0933-1433
• DOI: 10.1007/bf00160368

✦ Synopsis

Measures of Canadian fertility (total fertility rate and fifteen-year agespecific fertility rate F15_29 ) and relative cohort size (population aged 3 0 -6 4 years divided by population aged 1 5 -2 9 years) show a close co-movement between 1940 and 1976 but record a marked departure since then. The application of cointegration techniques to these series shows that they do not form an equilibrium relationship even over the period . Contrary to the expected relationship between relative cohort size and relative income, income data by age groups show that there is no tight relationship between them. The absence of an equilibrium relationship between relative cohort size and fertility, therefore, does not necessarily imply that Easterlin's hypothesis is false. * I would like to thank Paul Maxim for allowing me to use his data set for this analysis. My thanks are also due to Peter Smith and three anonymous referees for their constructive comments on this work.

54

T. Abeysinghe "relative numbers". The argument is that the relative income of young people falls as the number of young people increases relative to older people. The "relative cohort size" (RCS) of the two groups can be used, therefore, as a proxy for relative income. Easterlin measures RCS by the ratio of the male population aged 30-64 years to the male population aged 15-29 years. An upward movement of this ratio indicates an increase in relative income.

Although most of the research on Easterlin's hypothesis has been conducted using U.S. data, in a recent paper Wright (1989) applies a Granger causality test to assess the relationship between total fertility rate (TFR) and RCS in sixteen European countries. He finds little support for Easterlin's hypothesis. Wright, however, emphasises that his conclusion depends on the appropriateness of the RCS as a measure of relative income.

A number of authors have refuted Easterlin's hypothesis by examining the relationship between changes in fertility and changes in relative income measures (e.g., Wright 1989;Rutten and Higgs 1984;Butz and Ward 1979). In a reply to Rutten and Higgs, Easterlin (1984) questions the validity of such procedures by arguing, "The scatter diagrams of annual rates of change of fertility against relative economic status.., push both the hypothesis and data beyond the purpose for which they were intended" (p. 213). This is in fact a very important aspect to consider when testing strategies are formulated. Like many theories in economics, theories of fertility (Easterlin's hypothesis as well as the "new home economics") explain equilibrium relationships. The test procedures which utilize differenced data series essentially throw away this information and examine the short run disequilibria (Hendry 1986).

The recent advances in the econometric technique of "cointegration" offer a much better testing framework for equilibrium theories of economics. In fact, a cointegration test should be a precursor to Granger-causality tests such as those used by Macunovich and Easterlin (1988) and Wright (1989) since the differencing of cointegrated (equilibrium) relations produces noninvertible moving average (MA) processes which cannot be approximated by finite order autoregressions.

In this paper we use a cointegration approach to analyse the relationship between Canadian fertility and relative cohort size. We choose Canadian data for two reasons. Firstly, as Easterlin found with U.S. data a plot of Canadian TFR and RCS over the period 1940 to 1975 (the period covered by Easterlin 1980) undoubtedly suggests a close relationship between them (see Fig. 1). In fact, Easterlin and Condran (1976) have reached the conclusion that the movements of TFR and RCS in Australia, Canada, England and Wales and the United States support Easterlin's hypothesis. As opposed to this conclusion Wright and Maxim (1987) find, using a technique called age-period-cohort analysis, no relationship between Canadian fertility and RCS. It appears, therefore, that Canadian data provide a good testing ground for Easterlin's hypothesis. Secondly, despite the availability of fertility and population data in Canada far too little research on Easterlin's hypothesis has been done using these data. Perhaps the parallel movement of fertility in the United States and Canada may have led to believe that the research findings in the United States are equally applicable to Canada. This paper is partly an effort to stimulate more research on Easterlin's hypothesis using Canadian data.

The conclusion that we reach in this paper is that Canadian fertility and relative cohort size do not form an equilibrium relationship. A lack of relationship is quite obvious graphically since mid seventies (Figs. 1 and2). The cointegra-