Too many, too fast? Dynamics of net migration in OECD, 1984–2001

Abstract

A dynamic model of migration developed by Hatton (1995) has been applied to the panel of 23 OECD countries observed during 1984–2001. Migration flows have been found to have a tendency to overreact to changes in economic conditions. Thus, simulations have shown that in the Anglo-American group of countries (Australia, Canada, Ireland, UK and USA) a given relative improvement in economic circumstances which brings an extra 0.840 immigrants per 1000 population per year (334 800 in total) in the short run, brings somewhat fewer (288 700 in total) in the long-run.

Notes

¹ For anecdotal evidence see, for instance, The Economist, ‘Half a Billion Americans?’, 22 August 2002; The Economist, ‘Where The East Europeans Go?’, 24 February 2005.

² The net migration rate is estimated using aggregate population data, as the difference between the total population growth and the natural population growth (the birth rate net of the death rate). This standard method avoids inconsistencies in migration accounting in different countries.

³ To see this, consider estimation of an error-correction model with two explanatory variables:

(i)

The usual way of estimating it is converting it into a linear auto-regressive distributed lag (ADL) model,

and apply a consistent estimator to it. Then one could reparameterize the estimated ADL into an ECM:

(ii)

If, however, one proceeds by estimating

(iii)

and then the original error-correction model with the residuals from (iii) in it, the implicit assumption is that the respective coefficients in (i) and (ii) should be equal. There is no problem with it, at least from the point of view of econometric theory, when variables are nonstationary in levels, as any correlation between the differences and levels of nonstationary variables can only be spurious. Therefore, estimating Equation iii(iii) separately will not produce systematically biased slope coefficients.

⁴ Full set of results is available from the author.

Table 2. Estimation results for 1–3