I’ve been playing around with a model of immigrant selection in which selection is a function of both the National IQs of the immigrants’ country of origin and the educational selectivity of emigrants relative to the various non-emigrating populations. The first part is based on the idea that emigrating is an IQ test with a selection ratio below 100%. That is, it is a test in which not all who desire to emigrate pass. If so, given national differences in true latent IQ, there should be “adverse impact” for emigrating, just as there is for tests when populations differ in IQ. That is, emigrating should be, on average, less problematic on the account of IQ alone for individuals from high IQ nations than low. This entails that more of the latter will be discriminated against and, as such, that those of the latter who do emigrate will be more selected, relative to the non-emigrating populations, than those of the former. If this is the case, then estimated selectivity will require good National IQ estimates. Of the National IQs provided by L & V and others, those from the Caribbean are of the worst quality (as indexed by L & V’s quality of data measure; refer to page 17 of L & V (2012) for a discussion of this measure).
A such, I tried to create better Caribbean estimates. Below, row A gives the country, B gives the final estimated IQ, C the quality of measures, D the point range for those countries which had multiple measures, E Lynn’s IQs, F Lynn’s IQ sample sizes, G the quality of the samples, H Lynn’s A(chievement)Qs, I the tests these were based on, J the quality of these samples, K my IQs, L the quality of these sample, M my AQs, N the tests these were based on, O the quality of these samples, P AQs based on CXC pass rates, Q the quality of these samples. Final IQ estimates were based on IQ and achievement estimated weighted for quality. For quality with respect to IQ I assigned 0 to samples if the total sample was below 100, 1 if the total sample ranged from 100-1000, 2 if the total ranged from 1000 to 2000 and so on. For quality with respect to AQs, I assigned 2 points to major international tests (e.g., PISA, PIRLS, SERCE) and 1 point to other tests.
As for the AQ scores I derived I used international standard deviations for PISA, PIRLS, SERCE, and LLECE and the national standard deviations for NAEP. I adjusted the NAEP scores down two points as, on IQ tests, the national US average is 2 points below the national UK average. The CXC scores were derived from math and reading pass rates based on the 2000 and 2001 CSEC, the Caribbean Secondary Education Certificate. These scores are probably not the best indexes of IQ but they are likely more reliable than Lynn’s connivance sample IQs. Following a modified version of L & V’s method, the relative difference between Caribbean nations on the CSEC was first calculated and then all scores were adjusted down by 19 IQ points. The 19 point adjustment was for the average underperformance of Belize, the Dominican Republic, and Trinidad and Tobaga on international tests relative to the UK AQ. My IQ estimate for the Bahamas was based on Johnson and Holmes (1987). An IQ of 103 was given for the 75th percentile of representative samples of secondary students in 1987 based on the Otis. This was adjusted 0.68 SD down, as the 75th percentile is that much above the mean, and 1.5 points were deducted for the Flynn effect, on the assumption that the 1982 Otis norms were used.
Many of these estimates are improbably low – for example, the Jamaican national IQ of 72.6. Until more reliable scores are found, the estimates presented should be taken as very rough approximations. Data is, in principle, available to make better estimates. For example, Jamaica has conducted a couple of nationally representative surveys in the last two decades which contain PPVT scores as an index of cognitive abilities. And the Netherlands-Antilles took part in the PISA 2009 survey. Finding published results though, with means and standard deviations, has been challenging. And publicly accessible data sets have not been released for these surveys. But such information, is probably floating around somewhere or could be made available on request.