survey data on neonatal, post neonatal and child mortality rates, the latter being for deaths occurring after the child’s first birthday, but before their fifth. Figure 2 shows the differences between survey based estimates and official data for eight countries. In all countries except Georgia, the difference between survey estimates and official data is greatest for neonatal deaths. This is not surprising, given the discussion above on misreporting infant deaths as stillbirths or miscarriages. What is more surprising is the finding that in six of the eight countries, the surveys appear to underestimate child deaths in comparison with official data. This finding is not easy to rationalize, except perhaps with reference to the misreporting of infant deaths as child deaths in the official data, thus boosting the child death count. One alternative explanation – that child deaths are misreported as infant deaths in the surveys – seems unlikely, for reasons discussed in Section 5. To summarise: there is little hard evidence on misreporting of infant deaths, but available data do suggest problems in all the countries of Caucasus and Central Asia, and also in Moldova, Russia and Ukraine (evidence relating to Belarus is scarce). Problems may also exist in some South East European countries, although evidence for these countries is mixed. Misreporting does not seem to be a problem in the EU accession countries of Central Europe and the Baltic states. 17 Figure 2: Survey estimates of neonatal, post neonatal and child mortality rates (per cent difference from official rates) Source: Official data: WHO Health for All Database, Survey data: Uzbekistan DHS 1996, Kyrgyzstan DHS 1997, Kazakhstan DHS 1999, Georgia and Romania RHS 1999, Armenia and Turkmenistan DHS 2000, and Azerbaijan RHS 2001. See Appendix I. It is important that the reader is not given the impression of systematic misreporting in these countries as a result of directives from ministries and high ranking bureaucrats. Certainly, incentive systems, coupled with the looser Soviet definition of live birth that gives room for some interpretation, may encourage a degree of misreporting. Equally, however, the trauma felt by medical staff who often have to care for dying infants with inadequate training and equipment, and who may feel powerless to save them, should not, perhaps, be underestimated.10 10 The authors are grateful to Iryna Mogilevkyna for this insight. -21 -Azerbaijan 1996-2000 Turkmenistan 1995-2000 Kyrgyzstan 1993-97 Kazakhstan 1994-99 Uzbekistan 1992-96 Georgia 1995-99 Romania 1995-99 Armenia 1995-2000 Neonatal Post Neonatal Child 18 4. RELIABILITY OF REGISTRATION The definition of a live birth, and its interpretation by medical staff, are not the only factors that can affect the official infant mortality count. As is noted in Section 2, only births and deaths that are registered are included in official statistics. In the Soviet Union and many other communist countries, parents were expected to register the births of children, except in the case of stillbirths and deaths that took place before the mother left the hospital, when it was the job of the hospital administration to make the declaration. In many countries, similar procedures exist today. To a large extent, therefore, the completeness of the registry will depend on parental willingness to register the birth and death of a child. Problems with the registration of births and deaths in CEE/CIS countries are under-researched, and hard evidence exists only for a small number of countries. Table 7 shows 2000 survey evidence of proportions of children at 0-5 months, 6-11 months and 48-59 months who were unregistered at time of interview in seven countries in the region. In Tajikistan non-registration is particularly acute: the births of four in ten children aged between 6 months and one year, and more than one in ten children aged between four and five years (48-59 months) were not registered in 2000. The situation in Georgia is also severe, with the births of 5 per cent of children aged 48-59 months not registered. The reason for high levels of non-registration in Tajikistan and in Georgia is simple – registration at the time of the survey was costly.11 In Albania, Bosnia-Herzegovina, Moldova and Uzbekistan, on the other hand, where registration was formally free, the picture looks more benign. Table 7: Births not registered 2000, by age of child (per cent) under 6 months 6-11 months 48-59 months 0 Source: MICS 2000 data. See Appendix I. Note: Information on birth registration in the MICS refers to proportions of children aged under 5 years whose birth has not been registered. Includes some cases where parents are unsure whether their children’s births are registered. 11 Formal birth registration fees have recently been abolished in Tajikistan, but persist in Georgia. In Georgia, moreover, birth registration fees for infants born to mothers who are not in a registered marriage are double the fees for infants born to married mothers (personal communication, UNICEF Georgia). 19 However, the data in Table 7 refer only to living children. The proportion of children who have died whose births (and deaths) are not registered may be higher still. Evidence from a trace of over 500 medical death certificates issued by hospitals and other health facilities for Armenian infants in 2000 shows that a quarter of deaths remained