respiratory illness. This does not affect the infant mortality count, and is not considered here. 11 The reader should note the difference between mis-reporting, which is a conscious act or an unintended error by medical staff, and non-registration, which is an act of omission, often by parents, but which can also reduce the infant mortality count. Issues surrounding registration are discussed in Section 4. Classification of infant deaths as stillbirths and miscarriages Because in all countries, a high proportion of infant deaths occurs within a short period after birth (that is, in the first weeks, and even days and hours), the infant mortality rate is often decomposed into early-neonatal mortality (infant dies within 6 days of birth), neonatal mortality (infant dies within 27 days after birth) and postneonatal mortality (28 to 364 days). In general, one might expect the relationship between neonatal and post-neonatal mortality rates across countries to be reasonably stable. That is, countries with similar levels of neonatal mortality should also have similar levels of post-neonatal mortality. Velkoff and Miller (1995) find that for years up to 1990, in several republics of the Soviet Union, the ratio of official post-neonatal mortality rates to neonatal mortality rates was high when compared with Western countries. Is this still the case in CIS countries? Is it the case in other countries of Central and Eastern Europe? Table 5 shows ratios of post-neonatal to neo-natal mortality rates for 15 western European countries, and 19 countries in the CEE/CIS region. These are average ratios for available years between 1970 and 2000 when the infant mortality rate was between 10 and 20 per 1,000 live births. Countries are grouped by geographical region. Data for Western European countries set a benchmark: the sorts of expected ratios where infant mortality is accurately captured by official statistics: these range from 0.24 in Sweden (denoting 24 post neonatal deaths for every 100 neonatal deaths) to 0.64 in the UK, with an average ratio of 0.43. Average ratios in seven Central European and Baltic countries are only slightly higher, at 0.50. In the other three sub-regions on Table 5, ratios for most countries are considerably higher. Exceptions are Bosnia-Herzegovina, Croatia, FYR Macedonia and Georgia, where ratios are similar to those in Western Europe. In the other countries (Albania, Romania, Moldova, Russia, Ukraine, Armenia and Azerbaijan), given the official count of post neonatal deaths, the number of neonatal deaths appears to be understated.9 9 The alternative interpretation, that the number of post-neonatal deaths is overstated, seems unlikely. 12 Table 5: Average Ratios of Post neonatal to neonatal mortality rates in official statistics, European and Central Asian countries, 1970-2000 Western Europe Central Europe and Baltic states South Eastern Europe Western CIS Caucasus and Central Asia Average 0.43 Average 0.50 Average 0.84 Average 0.83 Average 1.76 Austria (10) 0.56 Czech Republic (5) 0.45 Albania (4) 1.80 Moldova (6) 0.96 Armenia (10) 0.85 Belgium (12) 0.51 Estonia (4) 0.57 BosniaHerzegovina (4) 0.45 Russia (8) 0.69 Georgia (7) 0.51 Denmark (7) 0.31 Hungary (12) 0.39 Bulgaria (21) 0.85 Ukraine (7) 0.85 Azerbajjan (2) 3.93 Finland (5) 0.26 Latvia (13) 0.69 Croatia (10) 0.39 France (11) 0.60 Lithuania (7) 0.56 FYR Macedonia (4) 0.45 Greece (11) 0.34 Poland (15) 0.40 Romania (2) 1.10 Ireland (8) 0.57 Slovenia (3) 0.48 Italy (11) 0.29 Netherlands (6) 0.37 Norway (7) 0.41 Portugal (10) 0.49 Spain (9) 0.45 Sweden (3) 0.24 Switzerland (7) 0.45 United Kingdom (5) 0.64 Source: WHO Health for All Database Notes: Average ratios (PNNMR/NNMR) are calculated for each country for years between 1970 and 2000 when official infant mortality rates were between 10 and 20 per 1,000 live births. Number of years from which averages are calculated are in parentheses. The evidence presented on Table 5 is generally supported by a recent analysis by Kingkade and Sawyer (2001), where information on the monthly probabilities of dying at less than four months (that is, in the first month, second month, etc.), and between four and ten months, are used to estimate the ‘true’ infant mortality rate. Their method assumes that official rates for deaths that occur up to the fourth month in CEE/CIS countries may understate the true situation, while official rates for deaths between four and ten months in infancy are generally correct. The method also assumes that the pattern of deaths up to the fourth month can be estimated using historical US and German data from points in time when their infant mortality levels (between months 4 and 10) were similar to those in the different CEE/CIS countries (see also Kingkade and Arriaga, 1997). Table 6 presents their underestimation factors, which range from close to zero per cent in most central European countries, to 25 per cent and more in most countries of South Eastern Europe and the CIS. Belarus is a notable exception, with an adjustment factor of only 13 13 per cent, similar to that of Latvia. It is worth noting also that Kingkade and Sawyer’s analysis does suggest, in contrast to the evidence on Table 5, a degree of underestimation in official rates in both FYR Macedonia and Georgia. However, it is worth noting that underestimation factors for Azerbaijan, Georgia, Tajikistan and Turkmenistan are calculated from only one