Health Outcomes of Children in Northern Uganda: Does Current IDP Status Matter?

Transcription

1 Health Outcomes of Children in Northern Uganda: Does Current IDP Status Matter? Carlos Bozzoli and Tilman Brück Work in progress Bonn, 4 May 2009

2 Overview Motivation Literature Methodological approach Data Summary statistics Econometric results Discussion and policy implications 2

3 Motivation Health status of children has short and long-run implications for human capital formation Research question on internally displaced persons (IDP) How does IDP camp decongestion affect child health? IDP camps and health positive effect: provide health services negative effect: may affect health negatively through overcrowding hence ambiguous net effect of IDP camps on health Methodological challenge and innovation camp decongestion and child health may be endogenous hence combination of household survey data with conflict event data - conflict data explains decongestion - decongestion explains child health status Policy relevance How quickly should IDP camps be dissolved in post-conflict period? Who should receive priority health care (with limited funds)? 3

4 Health of Children During Conflict Guha-Sapir and van Panhuis (2002): survey effects of conflict on health: food shortages, poor sanitation, lack of shelter, less health care etc Ghobarah et al (2003): cross-country data war reduces disability-adjusted life years (DALYs), especially for young children Akresh and Verwimp (2006): household survey data mass violent conflict in Rwanda affects health outcomes in children (height for age) in poor and non-poor households Ssewanyana and Younger (2005): DHS data determinants of infant mortality in Uganda in 3 decades Research gap: HOW does war affect child health? 4

5 Violent Conflict in Northern Uganda Civil war in Northern Uganda LRA staging surprise attacks to loot and recruit fighters thus growing IDP population from 1996 from 2002, government forces population into camps geographic variation in conflict intensity August 2006: ceasefire - with continued insecurity slow process of decongestion of IDP population late 2005: 1.7 million people May 2007: 0.8 million people many IDPs commute between camp and home 5

6 Child Health in Northern Uganda Headcount poverty rates (household survey data) 1993: 74% in North versus 56% in Uganda 2006: 61% in North versus 31% in Uganda Under 5 mortality rates (DHS data) : 200 per 1000 (IDPs) versus 150 in Uganda (overall) due to malaria, diarrhoea and malnutrition in IDP camps (WHO) Under 5 morbidity rates (DHS data) fever in 2006: 60% of IDPs versus 41% in Uganda diarrhoea in 2006: 44% of IDPs versus 26% in Uganda Core immunization rates for young children (DHS data) 2006: 54% of IDPs versus 46% in Uganda Research gap: effect of remaining IDP on child health? 6

7 Summary of Negative Health Effect of Camp 7

8 Methodological Approach Challenge: simultaneous decisions IDP status and child health status may be jointly determined Data situation we only have household survey data available (which have a very high standard for a conflict-affected area) but there are also data available on the conflict Solution and methodological approach of the paper merge household data with conflict data 8

9 Measure Effect of IDP Status on Welfare Process war Estimated Model data B war displacement, esp. from 2002 displacement, esp. from 2002 stay and coping return and coping stay or return? coping welfare: e.g. child health data A welfare: e.g. child health Problem Potential simultaneity of location and health status Solution: estimation in 2 steps E1: war stay/return E2: stay/return welfare 9

10 Econometric Approach Postulate a recursive model with potentially endogenous dichotomous variable y 1i * (IDP residence status of the child) and the variable of interest y 2i * (child health status - morbidity indicator) two variables are linked through the system (Maddala 1983): y y * 1i * 2i = β X = γy 1i 1i + u 1i + δ X Estimation y 1i * will be explained in part by conflict history of the household parameters are estimated via ML identification requirements discussed in Wilde (2000) small sample properties discussed in Fabbri et al (2004) 2i u 21 10

11 Household Survey Data: UBOS/FAFO 4000 households among IDP population and returnees in April/May 2007 Information on household members, including acute illnesses in last two weeks Survey is representative of households who have ever been displaced Pro: sample size (versus DHS); survey soon after end of war Con: no anthropometric information, calorie intake or health status; self-reported information Illness in the 2 Weeks Prior to Survey of Children under Age 10, by Age 30.00% 25.00% 20.00% Total Fever/Malaria Diarrhoea Cough/TB 15.00% 10.00% 5.00% 0.00%

12 Conflict Intensity Data: ACLED ACLED: Armed Conflict Location and Events Dataset (Uppsala/PRIO) 1276 village-level events in Uganda since 1962 (from books and press) 547 events in North in with 154 events in % battle events 8% rebel activity Intensity of LRA Events in Northern Uganda by Administrative Unit 12

13 Conflict Intensity Index Calculate conflict intensity index: C ( l) = g( d( c i, l)) weigh for distance: d( c, l) = i i c i l Index Conflict Score, selected locations Gulu Pader Lira Kampala g(x) = exp(-άx) 20 ά = 5 25 km = km = km = weigh for adjacent years: Year discount factor=

14 Children s General Characteristics Moved Away from Camp Still in Camp or Commutes p-value of Ho: Difference=0 Indicator: Father known to be alive 88.5% 84.7% Indicator: Mother known to be alive 95.2% 91.9% Age of Child (yr) Indicator: Access to safe water 52.7% 92.7% 0 Indicator: Mother known to be Widow 6.1% 5.0% Indicator: Female-Headed Household 15.8% 17.9% Number of Assets per HH Indicator: HH receives food 65.2% 81.3% 0 Indicator: HH receives seeds 51.3% 66.7% 0 Age of Head of HH 38.5% 38.3% Conflict Intensity, Birthplace (HH Head) Conflict Intensity, Location (HH Head) N 6,914 2,587 14

15 Children s Health Status Moved Away from Camp Still in the Camp or Commutes p-value of Ho: Difference=0 Joint Indicator: Fever-Malaria, Diarrhoea, Severe Cough-TB, or Vomiting in last 2 weeks 18.10% 18.50% Indicator: Fever-Malaria in last 2 weeks 10.0% 10.2% Indicator: Diarrhoea in last 2 weeks 3.3% 3.8% Indicator: Severe Cough-TB in last 2 weeks 2.4% 1.9% How healthy would children in camp be - had their parents moved? Health outcomes may also be driven by choice to stay/return Hence model this decision explicitly 15

16 Results: Is the Person still in/near IDP Camp? E1: Logit Model for Adults Aged Years Coefficients Standard Errors Age Age Sq * Female Ever did agriculture? Ever herded animals? ** Ever involved in petty trade? Distance to place of birth, Violence intensity at birthplace in *** Violence intensity at location in *** Constant Observations 4958 * significant 10%; ** significant 5%; *** significant 1% Higher conflict index: more likely to stay in camp 16

17 Results: Was Child Ill during the Last 2 Weeks? All Causes Fever/Malaria Diarrhoea Cough/TB Age Mother Age Mother Sq * Indic: Age Mother Unk Indic: Mother Alive -0.31* -0.51* Indic: Mother Life Status Unk Indic: Age <12 mo 0.436*** 0.455*** 0.56*** Indic: Age *** 0.515*** 0.729*** Indic: Age *** 0.54*** 0.478*** Indic: Age * Indic: Age Female *** ** *** Number of Assets Ind: camp resident/commuter 0.644* *** Dependency ratio ** * HH Size Sq HH Size Ind: Access to safe water ** ** Log Distance to Safe Water ** Age Head of HH 0.045** * Age Head of HH Sq 0** Indic: Female Head of HH Constant *** *** 17

18 Robustness checks Control for access to health facilities Control for conflict intensity in 2002 Control for facilities in camp (retrospective question for returnees) Use alternate weighting scheme to define conflict intensity District fixed-effects 18

19 Summary Child illness increases with/for maternal orphan-hood young age of child poor access to safe water being in/commuting to IDP camp (esp. TB: overcrowding?) Last result is in contrast to summary statistics our analysis: children have 60% higher risk of illness in camps Policy implications: focus health care on access to safe water in return sites on children in crowded IDP camps 19

20 Thank you for your attention! 20

21 Bonus Slides: Coping and IDP Status 21

22 Differences in Coping by Gender Activity Female Male P-val Difference Not Significant Any activity 0,936 0,935 0,873 Cultivates 0,89 0,883 0,282 Herding 0,153 0,201 0 Petty trade 0,127 0,076 0 Income Sources Female Male P-val Difference Not Significant No income 0,606 0,624 0,1 Salary work 0,045 0,161 0 Own family business 0,205 0,167 0,015 Sales goods 0,431 0,333 0 Remittances 0,06 0,061 0,918 Life will improve in the next year 0,772 0,797 0,34 Tables show individuals aged Years. 22

23 Differences in Coping by IDP Status Resident or Commuter Returnee P-val Difference Not Significant Age of head of household 42,43 42,99 0,434 Dependency ratio 1,226 1,237 0,02 Person is single 0,337 0,336 0,925 Person is female 0,502 0,51 0,614 Age 29,4 29,5 0,667 Literate (can read) 0,591 0,61 0,242 Any activity 0,923 0,965 0 Cultivates 0,876 0,909 0,028 Herding 0,132 0,27 0 Petty trade 0,1088 0,088 0,096 Does not earn income 0,603 0,642 0,036 Income from salary work 0,093 0,115 0,306 Income from own/family business 0,187 0,181 0,779 Income from sales of goods 0,413 0,323 0,005 Income from remittances 0,058 0,067 0,587 Life will improve in the next year 0,749 0,852 0 Number of assets 5,7 6,26 0,002 N Table shows individuals aged Years. 23

24 Second Round: Any Economic Activity? Ind: Person is head of household 0,436* Ind: Person is spouse of household head 0,299 Ind: Person is son/daughter of household head 0,087 Ind: Single -0,356* Ind: Female -0,026 Age (years) -0,006 Age squared (years) 0,000 Ind: Person knows how to read -0,057 Household size 0,094** Household size square -0,005** Dependency ratio -0,143*** Age of head of household -0,014 Age of head of household square 0,000 Indic: Female head of household 0,269 Indic: Head of household ever had animals 0,426*** Indic: Head of household ever involved on petty trade 0,123 Indic of violence in location ,004 Ind: IDP resident or commutes to IDP (instrumented) -1,478 District dummies (results not shown) Constant 2,673 Number of observations

25 Activities: Violence Reduces Cultivation Mg/Incremental effects after bivariate probit on each outcome 0,15 0,1 0,05 0 Any Activity Cultivates Herding Petty Trade -0,05-0,1 ** -0,15 IDP or near Female Violence Intensity where HH located in 2006 (incremental effect of 1 interquantile range) Literacy Head of HH 25

26 Income Sources: IDP Status Lowers Salary Work Mg/Incremental effects after bivariate probit on each outcome 0,4 0,3 0,2 *** 0,1 0 No income Income Own/Fam Business Sales Goods Remittances Salary work -0,1-0,2-0,3-0,4 *** IDP or near Female Violence Intensity where HH located in 2006 (incremental effect of 1 interquantile range) Literacy Head of HH 26

27 Expectations: IDP Status Has Negative Effect 0,1 Mg/Incremental effects after bivariate probit on expectations 0,05 0 Next year better situation than now -0,05-0,1-0,15-0,2-0,25-0,3 *** IDP or near Female Violence Intensity where HH located in 2006 (incremental effect of 1 interquantile range) Literacy Head of HH 27

28 Summary IDP status has no significant effect on activity choices probably due to similarities in endowments of different locations other characteristics are more important policy implication: focus less on IDP status and decongestion Violence drives some activity choices especially reduction of cultivation hence reduce violence and strengthen security! IDP status does drive expectations hence focus further analysis on role of expectations for making decisions virtuous cycle of positive expectations and reconstruction? 28