STR loci are becoming increasingly important in forensic casework. In order to be used fairly and efficiently, the population genetics of these loci must be investigated and the implications for forensic inference assessed. A key population genetics parameter is the "coancestry coefficient", or FST, which is the correlation between two genes sampled from distinct individuals within a subpopulation. We present analyses of STR data, at geographic scales which range from national to regional, from the UK and other European sources. We implement a likelihood-based method of estimating FST, which has important advantages over alternative methods: it allows a range of plausible values to be assessed, rather than presenting a single point estimate, and it allows a subpopulation to be compared with a larger population from which a database has been drawn, which is the relevant comparison in forensic work. Our results suggest that values of FST appropriate to forensic applications in Europe are too large to be ignored. With appropriate allowance, however, it is possible to make use of STR evidence in a way which is efficient yet avoids overstatement of evidential strength.