Trends in child mortality must be seen against Iraq’s grim recent history (Table 1). Saddam Hussein became president in 1979. Two years later he launched an invasion of Iran – leading to eight years of war. Saddam Hussein’s regime mixed political repression with efforts to improve social welfare [Marr 2004]. And it seems that, despite the war, the general socio-economic and health conditions of Iraq’s population did not deteriorate. Nevertheless, at the war’s end the country was deep in debt, and this was one factor behind the attack on Kuwait just two years later [Towle 2005:172].

The invasion of Kuwait in August 1990 led immediately to strict economic sanctions on Iraq imposed by the UN Security Council (UNSC). Because about 70 per cent of Iraq’s food was imported, the government reacted quickly by instigating a system of household food rationing – the Public Distribution System (PDS). The five months before the start of the first Gulf war was a time of mounting food shortages, increasing prices, and rising rates of malnutrition [Harvard Study Team 1991]. The war itself lasted for just a few weeks, but especially through the aerial bombing campaign it badly affected Iraq’s infrastructure (e g, transport, electricity, water). Following the ceasefire, Saddam Hussein’s regime faced major uprisings – among Kurds in the north, and Shiites in the south. The latter rebellion was suppressed. But fighting with Kurdish forces continued, until in late 1992 the army withdrew completely from the three northern governorates, leaving them to administer themselves. That said, the sanctions originally imposed under UNSC Resolution 661 (UNSCR 661) remained in place. The Iraqi economy was very badly affected because the sanctions hit at its core – i e, the capacity to export oil in exchange for a wide range of crucial imports. The UNSC set as the main condition for the lifting of sanctions that Iraq should scrap its weapons of mass destruction.

Reports about the hard living conditions that were being experienced by the Iraqi people were common following the first Gulf war [Harvard Study Team 1991; Drèze and Gazdar 1991]. But information about the extent and depth of the suffering increased in the years that followed. It was clear that, on top of the damage done by the war, all basic aspects of life – food, health, water, sewage, employment, etc – were being adversely affected by the operation of the sanctions. There was mounting international debate as to who was responsible – Saddam Hussein’s regime, or governments like those of the US and UK that most supported the sanctions. These developments led the UNSC to pass Resolution 986 in April 1995. It was intended to ameliorate the situation by allowing Iraq to sell oil so that it could import basic supplies – especially of food and medicine – under UN supervision. But the process of negotiating with the Iraqi government and implementing UNSCR 986 was protracted. The first supplies of food and medicine delivered under the OFFP only arrived after a delay of about two years. By early 1998, however, there were some signs of improvement. Thus Garfield (1999:4) observed that “[s]ince March 1998 the oil for food programme has greatly increased access to essential supplies and the mortality rate [of children under five] has surely declined, but data are not yet available to estimate the magnitude of that decline.”

The period 1998-2001 appears to have seen some limited improvement in living conditions. Due to the OFFP, the amount of food distributed each month to households through the PDS

Note: This list draws from many sources and is not meant to be comprehensive.

increased somewhat. It was made easier for Iraq to import equipment so that it could rehabilitate its oil-producing capacity. Saddam Hussein’s regime also became more adept at circumventing the sanctions, e g, through covert oil sales. There were US/UK air strikes against military targets. UN weapons inspectors came and went – the renewal of their access by the regime at times resembling a bargaining chip that enabled some slight further easing of the sanctions [Falk 2004:27].

The situation might have evolved gradually. But the attack on the US on September 11, 2001 led to its transformation. In January 2002 president George W Bush singled out Iraq as part of an “axis of evil”. “Regime change” in Iraq became a leading topic of public debate in the US [Marr 2004]. The year 2002 involved increasing US preparations for an invasion, and a tightening of the imposition of the sanctions. World Bank estimates suggest that per capita GDP fell from 1,068 US dollars in 2000 to 743 dollars in 2002 [World Bank 2006]. The US/UK invasion – the second Gulf war – took place during March/April 2003. On May 1 president Bush declared that major combat operations were ended. And the UNSC lifted the economic sanctions later that same month (Table 1). Nevertheless, given the invasion and upheaval that followed in its wake, 2003 was perhaps an even harder year for the Iraqi people than those that preceded it – estimated per capita GDP fell to 479 dollars [World Bank 2006].

Clearly, the situation in Iraq has been extremely hard and unsettled in the years since 2003. The ILCS of 2004 revealed a very troubled picture. Whereas in 1980 Iraq had enjoyed living standards similar to the best in west Asia, by 2004 it had fallen far behind. And most Iraqi households remained heavily dependent on the PDS for their food supplies [UNDP 2005b:63].

As already noted, both the UNICEF survey and the ILCS collected birth histories. These can provide comparable and detailed data on fertility and child mortality, and are our point of departure here.

The UNICEF survey of 1999 was conducted in two parts. The first was undertaken in February/March in the south and centre of Iraq, i e, the area then still under the control of the government, with about 80 per cent of the population. The second part was conducted in April/May in the autonomous northern region, with around 20 per cent of the population. The respective sample sizes were 24,000 and 14,000 households. Considerable effort was spent on the training of field staff. Minor difficulties arose during the fieldwork which temporarily inhibited access to some areas, but these problems were resolved. Importantly, the UNICEF survey was focused specifically on estimating levels and trends in child mortality – with little additional information being gathered. The questionnaire used relevant parts of the Demographic and Health Surveys (DHS) core questionnaire. And, crucially, ever-married women aged 15-49 living in the sample households were asked questions on the number of children that they had ever borne and the number of these children that were surviving (i e, CEB/CS questions) before they were asked for their detailed birth history. In the country’s south and centre all the trained interviewers were women; the figure in the north was 80 per cent. Finally, an independent expert review panel was established to scrutinise the resulting data. It found no evidence of manipulation. Instead, it concluded that the data were of high quality – with no suggestion of deficiencies that sometimes occur in such material, like omission in the reporting of births and child deaths [UNICEF 1999; Ali et al 2003].

The annual infant (i e, under-1) and child (i e, under-5) mortality rates in Table 2 are estimates of the probability of dying between birth and the first and fifth birthdays respectively, expressed per thousand live births. Several points arise from comparing these rates. First, in relation to both the UNICEF and ILCS estimates, it is clear that most child mortality is accounted for by infant mortality, i e, mortality in the first year of life (at age zero). Only a modest proportion occurs at ages 1-4. Second, and relatedly, the trend in child mortality is dominated by the trend in infant mortality. Again, this is true for both sets of estimates. Third, the ILCS mortality rates are much lower than the UNICEF rates. Thus whereas only a modest adjustment is required to bring the TFRs of the ILCS up to those of the UNICEF survey, the corresponding adjustment needed in relation to child mortality is far greater. A strong implication – reflected in the ILCS report (UNDP 2005b:50-51) – is that, despite having re-collected all the birth histories, the resulting data are still very deficient in the reporting of child deaths. Fourth, while they generally increase between 1990 and 2003, the ILCS mortality rates do not show an abrupt rise in 1991. In contrast, however, the UNICEF under-5 death rate increases from about 66 per thousand in 1990 to around 118 in 1991. Moreover, if the UNICEF under-5 mortality rates are decomposed into infant mortality rates and mortality rates at ages 1-4, then only a small rise is indicated for ages 1-4. Put differently, most of the rise in child mortality indicated for 1991-98 is due to a rise in infant mortality.

Several questions arise from these comparisons. Is the rise in mortality in 1991 indicated by the UNICEF data real? If it is, then why is it not reflected in the ILCS data? And why was the rise largely restricted to infants? Finally, can the UNICEF and ILCS data be reconciled?

As noted, it is fanciful to argue that the UNICEF data were fiddled. But there is other evidence supporting a sharp rise in child mortality in 1991. The circumstances that affected the survival chances of children (e g, in relation to food, birth weight, nutrition, drinking water, prevalence of diarrhoea, etc) definitely underwent an abrupt deterioration at this time [Garfield 1999]. Child death rates derived from hospital data also suggest a sharp rise – although such rates are admittedly unrepresentative, difficult to calculate, and they could conceivably have been manipulated by the Iraqi government [Garfield 1999:34; Working Group 2005:135].

Most importantly, however, the International Study Team survey of 1991 provides independent support for a sharp rise in child mortality. Despite difficult working conditions, the IST survey appears to have been well done. All the interviewers were fluent

Table 2: Estimates of Fertility and Early Age Mortality from the UNICEF Survey and the ILCS

Notes: The UNICEF TFRs are quinquennial and have been allocated to the central years of the five-year periods to which they correspond most closely. The ILCSa (i e, adjusted) TFRs in column (iii) were pro-rated by 1.176 (see text). The infant and child mortality rates correspond to 1q0 and 5q0 values in life table terminology, and are expressed per thousand live births. As noted in the text, the ILCSa child mortality rates in column (xi) were adjusted first by 1.87 (for 1991 and later years) and second by 1.67; the corresponding adjustment factors for the ILCSa infant mortality rates in column

Sources: Columns (i), (iv) and (viii) Ali et al (2003:219); columns (ii), (v) and (ix) Pedersen (personal communication), but see also UNDP [2005b:49-52; 2005c:15].

in Arabic, and most were women [Ascherio et al 1992]. The survey collected what were in effect “partial” birth histories from adult women living in about 8,000 households. The histories were “partial” in that births occurring before January 1985 were excluded. The analysis related to 16,076 births and just 768 child deaths. The resulting estimates suggest that the under-5 mortality rate rose from about 43 child deaths per thousand live births during 1985-90 to 128 in 1991. The corresponding estimated rise in the infant mortality rate was from 33 to 93 [Ascherio et al 1992]. The UNICEF rates of Ali et al (2003) shown in Table 2 suggest that the IST survey underestimated child mortality during 1985-90. Thus the average UNICEF under-5 death rate for 1985-90 is 63 (compared to 43). This may help explain why the IST results imply a tripling in child mortality, rather than a near doubling. Nevertheless, when allowance is made for the much smaller sample size of the IST survey, the appreciably larger confidence intervals that attach to its estimates, and the particular period to which the IST rise in mortality pertains (i e, the first eight months of 1991, which includes the first Gulf war itself) then the two sets of estimates are reasonably compatible.

Yet if the abrupt rise in child mortality in 1991 is real, why is there no sign of it in the ILCS data?

The answer to this question probably lies in a hugely important feature of Iraqi life that began in September 1990 – the distribution to households of food rations (e g, cereals, pulses, salt) plus a few other basic supplies (e g, soap) through the PDS [Drèze and Gazdar 1991; FAO 1997; Garfield and Waldman 2003].

Recall that the PDS was established in response to the massive and abrupt drop in food availability that followed the imposition of sanctions. It has been extraordinarily important for the food security of most Iraqi households – especially the 80 per cent that remained under the control of the government. In the early years of the PDS most of the food that was distributed was grown in Iraq. But from 1997 onwards additional food became available through the OFFP. In most years the calorie content of the ration has been extremely limited [Working Group 2005:137-41]. For most of the 1990s it was barely enough to prevent starvation [Garfield and Waldman 2003:9].

The PDS works through a network of local agents, like grocery stores. Every household is entitled to a monthly ration for each of its members, and the ration has been the same irrespective of age or sex. The sole exception relates to infants – who from the start were entitled to a ration of infant formula (i e, breastmilk substitute) provided that the mother could produce a certificate stating that she was unable to breastfeed [Drèze and Gazdar 1991]. However, once a child reached its first birthday, and was no longer an infant, it was entitled to exactly the same ration as other household members. These circumstances provided a very strong incentive for the inclusion of new births on the ration cards. And an investigation in 2000 found that virtually all births were quickly registered in the rationing system [UNICEF 2002:8].

If the sharp rise in child mortality is real, then the failure of the ILCS death rates to show a similar abrupt rise around 1991 must be indicative of a sudden deterioration in the level of reporting of child deaths in the birth histories that were provided.

Recall that the working group concluded that (i) it was unlikely that there was a sharp rise in child mortality in 1991, (ii) any rise was modest – the group’s own estimates of under-5 mortality rates for 1986-90 and 1991-96 are respectively 85 and 95 per thousand, (iii) the Iraqi government might have tampered with the UNICEF data, and (iv) the ILCS data could be disregarded [Working Group 2005:50-55 and 129-36]. Enough has been said to cast considerable doubt on these conclusions. But recall too that in arriving at its estimates the Group placed most weight on its interpretation of CEB/CS data, especially from the 1997 Census.

As the working group acknowledges, there are several difficulties involved in translating CEB/CS data into estimates of child mortality levels and trends [e g, see United Nations 1990]. The CEB/CS questions are very simple. They provide no information on dates of birth or dates of death. Child mortality estimates based on the reports of younger women tend to be biased upwards due to the operation of various selection effects (e g, women aged 15-19 are young, mostly having first births, and both of these considerations involve an enhanced risk of child death). Estimates based on the reports of older women can be too low, since such women may be more likely to under-report dead children. Also, CEB/CS data aggregate births and deaths across very broad ranges of age and time. Thus the births and child deaths reported by women aged, say, 35-39 could have occurred over a wide range of ages and times – and this tends to obscure the occurrence of a discontinuity. Indeed, inasmuch as time enters into the analysis of CEB/CS data, it does so primarily through the ages that are reported by the women respondents. Furthermore, estimates of mortality from CEB/CS data involve selecting a particular model age pattern of child mortality, which is then assumed to remain constant. This means that it is hard to handle circumstances involving an abrupt alteration of age pattern – e g, a sharp rise in infant mortality, with little change at older ages.

The CEB/CS data and attendant mortality estimates summarised in Table 3 suffer from all of these difficulties. Since there is reason to believe that the age pattern of mortality changed around 1991, different patterns were used in deriving the child mortality estimates and dates from the 1987 and 1997 data. But this brings little benefit – given the magnitude of the other difficulties. Notice, in particular, the very large rises in child mortality indicated by the reports of younger women (especially those aged 20-24 and 15-19) in both 1987 and 1997. These rises are certainly spurious and reflect the operation of strong selection effects.

However, there is another difficulty with the CEB/CS data from the 1997 Census. This relates to something that the working group is much concerned with – since it makes a “huge difference” to the resulting estimates [Working Group 2005:130]. But it is something that the group fails to explain. The problem is that there was: “a substantial level of non-response, not to the question about the number of children ever born, but to the question about the number of children surviving. The treatment of this nonresponse has a considerable effect on the estimates of child mortality derived from the data” [Working Group 2005:51]. The difficulty is much greater for 1997 than for 1987. Thus whereas in 1987 only 1.5 per cent of women failed to answer the question on surviving children, in 1997 this figure rose markedly to 3.5 per cent. The working group’s analysis of the 1997 data leads to an appreciably higher estimate of child mortality in the early 1990s, if it is assumed that those women who did not respond were those whose children had died. But the group regards this possibility as unlikely, and instead favours the view that most of the women who did not respond had lost no children. To be specific, they suggest that “for women none of whose children had died the interviewer simply put a dash in the relevant box … under this assumption most of the ‘missings’ would have lost no children” [Working Group 2005:130]. This helps to explain why the working group’s estimate of the rise in child mortality is so modest.

Yet, of course, there is a ready explanation for why Iraqi women were less inclined to answer the census question on the number of surviving children in 1997. In all likelihood, many of the women who did not respond were precisely those who had lost a child in the previous few years. But the names of these dead children were still on the ration cards. Moreover, for the same reason it appears probable that – analogous to what seems to have happened in the ILCS – some of the women who did answer the question inflated their number of surviving children (i e, failed to report fully on those that had died). In short, the 1997 CEB/CS Census data probably reflect a systematic tendency to overstate child survivorship. Indeed, the substantially increased level of non-response to the 1997 Census question on surviving children, as revealed by the working group, constitutes further support that the food rationing system produced a downward bias in the reporting of child deaths.

It remains to assemble an integrated account of child mortality. The UNICEF under-5 mortality rates derived by Ali and colleagues provide the benchmark. It is possible that these rates underestimate child mortality a little – e g, for the 1970s and 1980s – but they are unlikely to be overestimates. And even for the 1970s and 1980s the UNICEF figures seem reliable. Thus, for years before 1990, comparison with estimates from earlier surveys shows the UNICEF rates “to be in reasonably good

Age 1987 1997 Group CEB CS Date Under-5 CEB CS Date Under-5 Mortality Mortality Rate Rate

15-19 0.170 0.150 1986.7 176.5 0.118 0.103 1996.7 177.1 20-24 1.088 0.989 1985.5 112.8 0.815 0.728 1995.6 128.2 25-29 2.573 2.375 1983.8 84.8 2.031 1.834 1993.9 105.7 30-34 4.263 3.929 1981.8 80.5 3.491 3.167 1991.9 95.4 35-39 5.785 5.284 1979.5 83.0 4.825 4.393 1989.6 88.1 40-44 6.460 5.807 1977.0 90.1 5.887 5.296 1987.0 92.9 45-49 6.715 5.879 1974.1 100.8 6.482 5.787 1983.8 91.8

Notes: For both censuses the Working Group gives CEB/CS data that are both unadjusted and adjusted for non-response. However, given the present argument, and reasons of space, the 1987 data above are those adjusted, while the 1997 data are unadjusted. The child mortality rates, and dates, above were obtained using standard estimation techniques [United Nations 1990]. Those for 1987 were derived assuming a west model age pattern, while those for 1997 assumed an east model; this was done because the sharp rise in infant mortality compared to mortality at ages 1-4 suggests a move from a west to an east model [Ali et al 2003: 221; Working Group 2005:154]. However, it should be stressed that these data, especially for 1997, have major limitations as indicators of child mortality.

Source: CEB/CS figures from Working Group (2005:147).

agreement with those from other sources” [Ali et al 2003:222]. And a notable feature of Table 3 is that the under-5 mortality estimates derived from women aged 30-49 in 1987 are close to the UNICEF rates in Table 2.

With this as background, the ILCS under-5 mortality rates in Table 2 have been adjusted in two steps. The argument here has been that the logic of the PDS means that women were particularly reluctant to report child deaths if they had occurred in 1991 or more recently. It follows that, starting in 1991, there was an abrupt and sustained fall in the level of reporting of child deaths in the ILCS birth histories. Thus, whereas during 1980-90 the ILCS rates were 59.66 per cent of the UNICEF rates, during 1991-98 the average level fell to 31.90 per cent (Table 2). Accordingly, starting from 1991, the ILCS rates must be adjusted upwards by

1.87 (i e, 59.66/31.90) to correct for an abrupt decline in the level of death reporting. The ILCS (first adjustment) series in the figure is the result. Although this first adjustment is sizeable in proportional terms, it is modest in absolute terms, and leaves the adjusted figures well below those of the UNICEF rates. This brings us to the second adjustment – which is needed because, for the several reasons mentioned, the general level of reporting of child deaths in the ILCS was poor. Again, the UNICEF under-5 mortality rates for 1980-90 provide the benchmark. They imply that a further upward adjustment of 1.67 (i e, 100/59.66) is required. The ILCSa series in the figure shows the result after this second adjustment. Analogous adjustments were made for infant mortality. Table 2 gives both ILCSa mortality series.

20 40 60 80 1970 1975 1980 1985 1990 1995 2000 2005 Year ILCS (first adjustment)UNICEF ILCSa

Table 4: Estimates of Excess Child Deaths in Iraq under Three Counterfactuals

Notes: For the approach of producing estimates of child deaths per month, see Working Group (2005:55).The child mortality rates for years 1999-2003 used in framing the second counterfactual were respectively: 40.2, 38.6, 37.0, 35.5 and 34.0 per thousand; the corresponding infant mortality rates were 33.5, 32.2, 30.8, 29.6 and 28.3 per thousand [see also Ali et al 2003:223]. The simple and illustrative nature of this exercise is reflected, for example, in the fact that the number of births stays the same in each set of calculations.

Discussion

The account of child mortality in Iraq that emerges from this work is quite clear, detailed, and well supported. The figure suggests that child mortality declined quite rapidly in the 1970s. In the middle of that decade the child death rate was either around or a little above 100 per thousand, but by the end of the decade it had fallen to about 80. However, the pace of decline slowed in the 1980s. Both the UNICEF and ILCS data suggest that such a slowdown occurred – and it probably reflected the Iran/Iraq war begun by Saddam Hussein. There is no good reason to doubt the estimates of Ali et al (2003) that, before the invasion of Kuwait, the child death rate was around 60-65. Nor are there convincing reasons to doubt that the first Gulf war, and the events that surrounded it, combined to produce a near doubling of the child mortality rate in 1991. The main component of this rise was a rise in infant mortality. And, among other things, a deterioration in breastfeeding practices probably played a significant part. Both surveys then suggest that, with the UN sanctions in full force, the child mortality rate rose further in the 1990s. People’s assets were run down, and the sanctions probably had a cumulative weakening effect. Both the UNICEF and ILCS data suggest that the highest under-5 death rate in the 1990s happened in 1998.

The ILCSa rates suggest that under-5 mortality fell from about 138 per thousand in 1998 to 107 in 2001 – consistent with some recovery in living conditions. Of course, these estimates are tentative and rough. Nevertheless it seems possible that child mortality in 2001 was lower than in any year since 1991. The fall occurs when the UN oil for food programme seems at last to have had an effect – recall Garfield’s observation that this was so. The working group (1995:178) judged that it could not conclude anything about the effects of the OFFP on under-5 mortality. But that is not the position taken here. With other developments, such as increased circumvention of the sanctions, the OFFP may well have had a beneficial effect in the late 1990s.

Finally, in this account, the ILCS data suggest that the child death rate rose in 2002, and that it rose further in 2003 (in the figure). This is plausible, given what we know about events on the ground. The ILCSa mortality rate of 144 for the invasion year of 2003 is higher than for any year since 1980. This is unlikely to be mere coincidence.

Previous estimates of excess child deaths derived from direct data on mortality have varied widely. The IST estimate of 46,900 for the first eight months of 1991 was a reasonable first attempt. But it pertained to a special time, and was probably too high

– since it underestimated mortality during its baseline period (i e, 1985-90). On the other hand, the working group’s estimated range of 45,000 to 68,000 excess child deaths for 1991-96 is unjustly low. The group’s analysis of child mortality pays scant attention to the conditions from which the data originate, and it is probably based on a misreading of, and over-reliance on, the 1997 Census CEB/CS data. There are no persuasive reasons to question the general range of excess child deaths estimated from the UNICEF data by Ali and colleagues – namely, somewhere between 3,80,000 and 4,80,000 for 1991-98. Inasmuch as criticism of these numbers is reasonable, it is that the higher number of the range arises from a counterfactual that embodies a long run rate of decline in mortality that may be a little too great. Finally, the present work has extended the time period to 2003. And in developing its counterfactuals it has tried to take some account of the slowing of the rate of child mortality decline in the 1980s. The resulting estimated range of excess under-5 deaths for the entire period 1991-2003 is between about 670 and 880 thousand, with a central figure of 770 thousand. The crude nature of these figures should require no emphasis. But there has clearly been a very large number of excess child deaths in Iraq.

In concluding, one criticism of the present exercise might be to question part of its point. Estimates of excess deaths are usually made with respect to events that are relatively short and discrete (e g, famines). However, in the case of the greatly raised level of child mortality in Iraq in the years that have followed the invasion of Kuwait, there is no obvious end in sight. Therefore, fair comment on any exercise such as this is that the resulting estimate of excess deaths will almost certainly rise as year follows year. Perhaps a better way of describing things is to say that the country’s mortality trajectory has been shifted onto an entirely new and tragic plane.

m

Email: t.dyson@lse.ac.uk

[I thank Athar Hussain, Mike Murphy, Yahia Said, and Jon Pedersen foradvice. The usual disclaimer applies.]

Ali, M, J Blacker and G Jones (2003): ‘Annual Mortality Rates and Excess

57(2):217-26. Ascherio, A, R Chase, T Coté, G Dehaes, E Hoskins, J Laaouej, M Passey,

Barringer, F (2003): ‘Fewer Iraqi Men: Dead or Undercounted?’, New York Times, August 8.

Bhatia, B, M Kawar and M Shahin (1994): ‘Unheard Voices: Iraqi Women on War and Sanctions’ in S Bloom, J Miller, J Warner and P Winkler (eds), Hidden Casualties: Environmental, Health and Political Consequences of the Persian Gulf War, Earthscan, London.

Drèze, J and H Gazdar (1991): ‘Hunger and Poverty in Iraq’, World Development, 20(7):921-45.

Dyson, T (2006): ‘On the Death Toll in Iraq since 1990’, Working Paper, London School of Economics.

FAFO (2004a): ‘Iraq Multiple Indicator Rapid Assessment – Household Questionnaire’, available at www.fafo.no/ais/middeast/iraq/ IMRA%20Household%20English.pdf. Accessed May 2006.

– (2004b): ‘Iraq Multiple Indicator Rapid Assessment – Women and ChildQuestionnaire’, available at www.fafo.no/ais/middeast/iraq/ IMRA%20W&C%20English.pdf. Accessed May 2006.

Falk, R (2004): ‘Iraq, the United States, and International Law: Beyond the Sanctions’ in T Ismael and W Haddad (eds), Iraq, The Human Cost ofHistory, Pluto Press, London.

FAO (1997): FAO/WFP Food Supply and Nutrition Assessment Mission toIraq, FAO Document Repository, Rome, available at www.fao.org// docrep/004/w6519e/w6519e00.htm. Accessed June 2006.

Garfield, R (1999): ‘Morbidity and Mortality among Iraqi Children from 1990 Through 1998: Assessing the Impact of the Gulf War and Economic Sanctions’, Campaign against Sanctions on Iraq (CASI) internet version, available at www.casi.org.uk/info/Garfield/dr-garfield.html.Accessed May 2006.

Garfield, R and R Waldman (2003): Review of Potential Interventions to Reduce Child Mortality in Iraq, Basics II Report, Office of Health and Nutrition of the Bureau for Global Health, US Agency for International Development, Washington DC.

Global Policy Forum (2004): ‘Ration Cards to be the Basis for ElectoralRegister’, available at www.globalpolicy.org/security/issues/iraq/election/ 2004/0607rationcards.htm. Accessed June 2006.

� �� � � � ��

REVIEW OF AGRICULTURE