Conclusions and suggestions for further work

We must be cautious about drawing general conclusions from this analysis as the sample was drawn from one locality. However, the extent to which migration behaviour with age can be explained by explanatory variables is likely to be informative about the process of migration.

We have identified three statistically significant peaks in migration behaviour with age during individuals' working lives; at just above age 20, at around age 35 and just below age 50. The size and location of the third peak has to be interpreted with caution as the data are sparse here.

We have shown that there is considerable heterogeneity in the population sampled, with a considerable proportion of individuals who are likely never to move.

The negative coefficient estimate for ldur indicates that the probability of migration decreases with duration of stay in the locality, consistent with the concept of cumulative inertia.

The simple logistic model takes no account of the fact that in a heterogeneous population, the individuals most likely to migrate are more and more underrepresented with increasing duration, and therefore inflates the duration of stay effect. To estimate the true effect of cumulative inertia, we must control for residual population heterogeneity.

For the years studied the likelihood of migration decreased with calendar time for the population surveyed.

The following time varying explanatory variables have been found to have a significant effect on migration (at the 10% level):

	Employment status
	Occupational status
	Promotion to service class
	First marriage
	Marital break-up
	Remarriage
	Presence of children age 15-16
	Marital status

It is evident that the third peak in the pattern of migration with age persist even after controlling for the time-varying explanatory variables. Remarriage appears to make a small contribution to this peak, however controlling for the presence of children of age 15-16 actually increases the size of the peak for those without children of this age.

The main effects model may be extended by the addition of interaction terms both between the time variables and between time and other explanatory variables. If these are confined to the linear term in age, there are 55 possible pairwise interactions. An interaction model has been fitted to this data by Borhani Haghighi and Davies (1999b). These throw light on questions such as:

1. Does the relative importance of the three peaks vary with calendar year?
2. Do patterns of migration behaviour for employed/self- employed/not working individuals relate to age?
3. Is the probability of migration after marriage break-up/remarriage age related?
We leave this for the student to explore.

As we have analysed migration data from only one locality, it is not clear how far the results are generally characteristic of the process of inter-county migration and how far they are location specific. Analysing datasets from some of the other SCELI localities would throw light on this question. See Davies and Flowerdew (1992) for some early comparative work.