3.1. Background Characteristics of the Study Participants
Table 1 presents the background characteristics of the study participants. The gender distribution among children showed a slightly higher proportion of females (51.19%) compared to males (48.81%). A considerable percentage of children (39.17%) were reported to have an average size at birth according to their mothers. Over 85% of the children were born at home, emphasizing a predominant prevalence of home births. The age distribution revealed that a similar proportion of children fell within the age groups of 12-36 months and 37-59 months during the survey, accounting for 44.5%. More than half of the children (54%) were breastfed for over a year, while close to 5% were not breastfed at all. A substantial majority (57%) of the children were born to mothers whose partners had no formal education. Notably, a significant proportion (74.35%) of the children were born to mothers without any formal education, with the remaining 25.65% born to mothers with at least primary level education. The majority of the children (90.06%) were born to mothers aged 19 years and above, with around 10% born to younger mothers (below 19 years old). About 55% of the children were born to mothers who were unemployed. Survey period distribution showed a slightly lower proportion (48%) of study participants surveyed during the 2000-2005 survey period.
At the household level, the distribution of children varied across wealth categories, with the largest proportion (31%) residing in the poorest households and the smallest proportion (13.65%) in the middle-income households. More than half of the children (55.09%) were part of households with more than six members. The majority of children (84.24%) lived in households headed by males, while those in female-headed households constituted 15.76%.
Regarding living conditions, a significant portion of children resided in households dependent on unimproved sources, with 64%, 86%, and 96% relying on unimproved drinking water, sanitation facilities, and solid cooking fuel, respectively, during the survey period.
At the community level, approximately 56% of children lived in established regions, while 36% were in emerging regions. The majority of children (85%) were situated in rural areas. These household and community-level characteristics provide a comprehensive overview of the socio-economic and environmental context in which the study participants were situated (Table 1).
Table 1: Background Characteristics of the Study Participants, Ethiopia, 2000-2019
Variables (N= 35,404)
|
N
|
%
|
Individual level drivers
|
Sex of child
|
|
|
Male
|
17282
|
48.81
|
Female
|
18122
|
51.19
|
Place of delivery
|
|
|
Home
|
30185
|
85.26
|
Health facility
|
5219
|
14.74
|
Child size at birth
|
|
|
Large
|
11140
|
31.47
|
Average
|
13867
|
39.17
|
Small
|
10397
|
28.37
|
Age of child
|
|
|
<1 year
|
3901
|
11.02
|
2-3 years
|
15762
|
44.52
|
>3 years
|
15741
|
44.46
|
Duration of breastfeeding
|
|
|
Never breastfed
|
1745
|
4.93
|
< 12 months
|
14455
|
40.83
|
12+ months
|
19204
|
54.24
|
Total children ever born
|
|
|
<5
|
20167
|
56.96
|
5+
|
15237
|
43.04
|
Partner’s education
|
|
|
No education
|
20373
|
57.54
|
Primary+
|
15031
|
42.46
|
Maternal education
|
|
|
No education
|
26323
|
74.35
|
Primary+
|
9081
|
25.65
|
Religion
|
|
|
Orthodox
|
12375
|
34.95
|
Muslim
|
15819
|
44.68
|
Others
|
7210
|
20.36
|
Maternal age at child birth
|
|
|
<19
|
3520
|
9.94
|
19+
|
31884
|
90.06
|
Maternal Employment
|
|
|
No
|
19355
|
54.67
|
Yes
|
16049
|
45.33
|
Survey period
|
|
|
2000-2005
|
17,027
|
48.09
|
2011-2016
|
18377
|
51.91
|
Household level factors
|
Asset-based wealth index
|
|
|
Poorest
|
10,977
|
31.00
|
Poorer
|
7161
|
20.23
|
Middle
|
4833
|
13.65
|
Richer
|
5790
|
16.35
|
Richest
|
6643
|
18.76
|
Household size
|
|
|
<6
|
15901
|
44.91
|
6+
|
19503
|
55.09
|
Sex of household head
|
|
|
Female
|
5579
|
15.76
|
Male
|
29825
|
84.24
|
Source of drinking water
|
|
|
Unimproved
|
22792
|
64.38
|
Improved
|
12612
|
35.62
|
Type of toilet facility
|
|
|
Unimproved
|
30337
|
85.69
|
Improved
|
5067
|
24.31
|
Type of cooking fuel
|
|
|
Solid fuel
|
34113
|
96.35
|
Clean fuel
|
1291
|
3.65
|
Community level drivers
|
Regional category
|
|
|
Established
|
19707
|
55.56
|
Emerging
|
12841
|
36.27
|
Urban
|
2856
|
8.07
|
Place of residence
|
|
|
Rural
|
30201
|
85.30
|
Urban
|
5203
|
14.70
|
Total
|
35404
|
100.00
|
Source: Authors analysis using EDHS 2000-2016
3.2. Bivariate Analysis Results
Table 2 reveals the associations between individual, household, and community-level explanatory variables and under-five mortality. At the individual level, several factors exhibited statistically significant associations (p<0.001) with under-five mortality. These include maternal education, religion, age at childbirth, employment status, partner's education, place of birth, total number of children ever born, duration of breastfeeding, child sex, age of the child, and child size at birth.
Among household-level variables, the household asset-based wealth index, household size, source of drinking water, type of sanitation facility, and type of cooking fuel demonstrated significant associations (p<0.001) with under-five mortality. However, the sex of the household head did not exhibit a statistically significant association (p<0.2) with under-five mortality.
Furthermore, at the community level, the regional category and place of residence were found to be significantly associated (p<0.001) with under-five mortality. These findings highlight the diverse array of factors at different levels that contribute to under-five mortality, emphasizing the multifaceted nature of this public health concern (Table 2).
Table 2: Bivariate analysis of explanatory variables by under-five mortality, Ethiopia, 2005-2016
Explanatory Variables
|
Under-five mortality (N=35,404)
|
Chi-square
|
Alive
|
Dead
|
N
|
%
|
N
|
%
|
Individual level drivers
|
|
|
|
|
Sex of child
|
|
|
|
|
34.18***
|
|
Male
|
15947
|
92.28
|
1335
|
7.72
|
|
|
Female
|
16406
|
90.53
|
1716
|
9.47
|
|
|
Place of delivery
|
|
|
|
|
59.80***
|
|
Home
|
27439
|
90.90
|
2746
|
9.10
|
|
|
Health facility
|
4914
|
94.16
|
305
|
5.84
|
|
|
Child size at birth
|
|
|
|
|
54.62***
|
|
Large
|
10017
|
89.92
|
1123
|
10.08
|
|
|
Average
|
12835
|
92.56
|
1032
|
7.44
|
|
|
Small
|
9501
|
91.38
|
896
|
8,62
|
|
|
Age of child
|
|
|
|
|
45.75***
|
|
<1 year
|
3645
|
93.44
|
256
|
6.56
|
|
|
2-3 years
|
14481
|
91.87
|
1281
|
8.13
|
|
|
>3 years
|
14227
|
90.38
|
1514
|
9.62
|
|
|
Duration of breastfeeding
|
|
|
|
|
6.30***
|
|
Never breast feed
|
689
|
39.48
|
1056
|
60.52
|
|
|
< 12 months
|
13749
|
95.12
|
706
|
4.88
|
|
|
12+ months
|
17915
|
93.29
|
1289
|
6.71
|
|
|
Total children ever born
|
|
|
|
|
37.42***
|
|
<5
|
18589
|
92.18
|
1578
|
7.82
|
|
|
5+
|
13764
|
90.33
|
1473
|
9.67
|
|
|
Partner’s education
|
|
|
|
|
69.98***
|
|
No education
|
18399
|
90.31
|
1974
|
9.69
|
|
|
Primary+
|
13954
|
92.83
|
1077
|
7.17
|
|
|
Maternal education
|
|
|
|
|
68.30***
|
|
No education
|
23864
|
90.66
|
2459
|
9.34
|
|
|
Primary+
|
8489
|
93.48
|
592
|
6.52
|
|
|
Religion
|
|
|
|
|
8.13**
|
|
Orthodox
|
11346
|
91.68
|
1029
|
8.32
|
|
|
Muslim
|
14382
|
90.92
|
1437
|
9.08
|
|
|
Others
|
6625
|
91.89
|
585
|
8.11
|
|
|
Maternal age at child birth
|
|
|
|
|
59.29***
|
|
<19
|
3095
|
87.93
|
425
|
12.07
|
|
|
19+
|
29258
|
91.76
|
2626
|
8.24
|
|
|
Maternal Employment
|
|
|
|
|
7.92***
|
|
No
|
17761
|
91.76
|
1594
|
8.24
|
|
|
Yes
|
14592
|
90.92
|
1457
|
9.08
|
|
|
Survey period
|
|
|
|
|
164.79***
|
|
2000-2005
|
15221
|
89.39
|
1806
|
10.61
|
|
|
2011-2016
|
17132
|
93.23
|
1245
|
6.77
|
|
|
Household level drivers
|
|
|
111.97***
|
|
Asset-based wealth index
|
|
|
|
|
|
|
Poorest
|
10052
|
91.57
|
925
|
8.43
|
|
|
Poorer
|
6382
|
89.12
|
779
|
10.88
|
|
|
Middle
|
4447
|
92.01
|
486
|
7.99
|
|
|
Richer
|
5232
|
90.36
|
558
|
9.64
|
|
|
Richest
|
6240
|
93.93
|
403
|
6.07
|
|
|
Household size
|
|
|
|
|
230.45***
|
|
<6
|
14132
|
88.87
|
1769
|
11.13
|
|
|
6+
|
18221
|
93.43
|
1282
|
6.57
|
|
|
Sex of household head
|
|
|
|
|
1.86
|
|
Male
|
27281
|
91.47
|
2544
|
8.53
|
|
|
Female
|
5072
|
90.91
|
507
|
9.09
|
|
|
Source of drinking water
|
|
|
|
|
40.47***
|
|
Unimproved
|
20667
|
90.68
|
2125
|
9.32
|
|
|
Improved
|
11686
|
92.66
|
926
|
7.34
|
|
|
Type of toilet facility
|
|
|
|
|
39.80***
|
|
Unimproved
|
27606
|
91.00
|
2731
|
9.00
|
|
|
Improved
|
4747
|
93.68
|
320
|
6,32
|
|
|
Type of cooking fuel
|
|
|
|
|
31.17***
|
|
Solid fuel
|
31118
|
91.22
|
2995
|
8.78
|
|
|
Clean fuel
|
1235
|
95.66
|
56
|
4.34
|
|
|
Community level drivers
|
|
|
|
|
|
|
Regional Category
|
|
|
|
|
19.28***
|
|
Established
|
17968
|
91.18
|
1739
|
8.82
|
|
|
Emerging
|
11712
|
91.21
|
1129
|
8.79
|
|
|
Urban
|
2673
|
93.59
|
183
|
6.41
|
|
|
Place of residence
|
|
|
|
|
54.00***
|
|
Rural
|
27461
|
90.93
|
2740
|
9.07
|
|
|
Urban
|
4892
|
94.02
|
311
|
5.98
|
|
|
*** p<.01, ** p<.05,
Source: Authors analysis using EDHS 2000-2016
3.3. Multilevel Analysis Results
Table 3 presents the results of a comprehensive multilevel binary logistic regression analysis, delving into the determinants of under-five mortality. The appropriateness of employing multilevel modeling was confirmed through a null model, revealing statistical significance (p<0.001). At the community level, children residing in rural areas exhibited a significantly higher likelihood of mortality (p<0.001) compared to their urban counterparts, emphasizing the contextual impact of the living environment.
Moving to household-level factors in Model III, several noteworthy findings emerged. Children in larger households, defined by at least six members, demonstrated a significantly heightened likelihood of mortality (p<0.001). Furthermore, households relying on improved sources of drinking water exhibited a significantly lower likelihood of under-five mortality (p<0.001). However, variables such as the sex of the household head, household asset-based wealth index, type of toilet facility, and type of cooking fuel were not found to be statistically significant predictors of under-five mortality.
Turning to individual-level factors, various determinants proved to be statistically significant in Model III. Female under-five children exhibited a significantly lower likelihood of mortality than their male counterparts (p<0.001). Additionally, children born with small and average sizes at birth faced a significantly higher likelihood of mortality compared to those born with a large size (p<0.001). The total number of children ever born to a mother was also a significant factor, with children born to mothers with more than five children facing a significantly higher likelihood of mortality (p<0.001). Partner's education emerged as a significant influence, as children born to mothers with uneducated partners faced a higher likelihood of mortality (p<0.001). Maternal education played a substantial role, with children born to uneducated, younger, and unemployed mothers facing a significantly higher likelihood of mortality compared to those born to educated, adult, and employed mothers (p<0.001). The survey period revealed a significant impact, indicating that under-five mortality was significantly higher during the period 2000-2005 compared to 2011-2016 (p<0.001). However, variables such as place of birth, child age in months, and maternal religion did not demonstrate statistically significant associations with under-five mortality in the multilevel logistic analysis.
These results underline the intricate interplay of diverse factors shaping under-five mortality, emphasizing the necessity for a comprehensive and nuanced approach to public health interventions aimed at addressing the needs of this vulnerable population.
Table 3 also reveals that the inclusion of individual-level predictors in the model has an impact on the variation in under-five mortality at the community and household levels. The effects of community variation increased from 0.011 in the null model to 0.017 in Model III, indicating a minimal variation in under-five mortality between regions and rural-urban settings. Similarly, household variation decreased from 0.271 in Model 0 to 0.233 in Model III, suggesting substantial variation between communities (regions and urban-rural areas) and households in under-five mortality.
Additionally, the estimated community-level Median Odds Ratio (MOR) increased from 1.239 in Model 0 to 1.301 in the final model, while the household-level MOR decreased from 2.891 to 2.623. This implies the existence of variation between households in under-five mortality in the country. The values of Akaike Information Criterion (AIC) and Bayesian Information Criterion (BIC) also decreased from 20580.7 and 20606.12 in Model 0 to 16522.83 and 16760.12 in Model III, respectively. The lower values of AIC and BIC in the final model indicate a better fit compared to the earlier models.
Furthermore, the likelihood ratio test (LR test) demonstrated statistically significant associations for the fitted multilevel models, affirming the appropriateness and good selection of the models employed for multilevel analysis. These findings collectively underscore the importance of considering various levels of influence to better understand and address under-five mortality in Ethiopia.
Table 3: Results of Multilevel Analysis of Under-five Mortality (N=35,404) Ethiopia, EDHS 2000-2019
Explanatory variables
|
Model 0
Coef. [CI]
|
Model I
Coef. [CI]
|
Model II
Coef. [CI]
|
Model III
|
|
Coef. [CI]
|
|
Community-level attributes
|
|
|
|
|
|
Region: Established(ref)
|
|
|
|
|
|
Emerging
|
|
.027[-.068, .122]
|
.032[-064, .127]
|
.036[-.084, .157)
|
|
Urban
|
|
-.159[-.359, .040]
|
-.041[-.246, .165]
|
-.012[-.251, .227]
|
|
Place of residence: Urban(ref)
|
|
|
|
|
|
Rural
|
|
.457[.306, .609] ***
|
.295[.117, .472] ***
|
.132[-.078, .034]
|
|
Household-level attributes
|
|
|
|
|
|
Household size: (<6) (ref)
|
|
|
|
|
|
Six and above
|
|
|
-.665[-.752, -.577] ***
|
-1.138[-1.261, -1.014] ***
|
|
Household head sex: (female) (ref)
|
|
|
|
|
|
Male
|
|
|
-.034[-.152, .084]
|
-.010 [-.141, .121]
|
|
Wealth status: Poor (ref)
|
|
|
|
|
|
Non-poor
|
|
|
-.061[-.158, .037]
|
-.079[-.189, -.032]
|
|
Source of drinking water: (unimproved)
|
|
|
|
|
|
Improved
|
|
|
-.159[-.264, -.054] ***
|
-.170[-.287, -.053] ***
|
|
Type of toilet: unimproved(ref)
|
|
|
|
|
|
Improved
|
|
|
-.162[-.318, -.006] **
|
-.142[-.317, .033]
|
|
Type of cooking fuel: Solid(ref)
|
|
|
|
|
|
Cleaned fuel
|
|
|
-.418[-.754, -.083] **
|
-.322 [-.717, .072]
|
|
Individual-level attributes
|
|
|
|
|
|
Sex of child: Male(ref)
|
|
|
|
|
|
Female
|
|
|
|
.197 [.104, .290] ***
|
|
Place of birth: Home (ref)
|
|
|
|
|
|
Health facility
|
|
|
|
-.126[-.316,.064]
|
|
Child age in months: (0-11months)
|
|
|
|
|
|
12-36 months
|
|
|
|
079[-.100, 2.57]
|
|
37-59 months
|
|
|
|
.109[-.083, .301]
|
|
Size at birth: Large(ref)
|
|
|
|
|
|
Average
|
|
|
|
-.291[-.402, -.179] ***
|
|
Small
|
|
|
|
-.229[-.347, -.111] ***
|
|
Duration of breastfeeding: Never(ref)
|
|
|
|
|
|
Less than 12 months
|
|
|
|
-4.118[-4.33, -3.91] ***
|
|
12 and above months
|
|
|
|
-3.74[-3.93, -3.55] ***
|
|
Total children ever born: (<5, ref)
|
|
|
|
|
|
5+
|
|
|
|
.832[.706, .959] ***
|
|
Partner’s education: Primary + (ref)
|
|
|
|
|
|
No education
|
|
|
|
.164[.049,.278] ***
|
|
Maternal education: Primary + (ref)
|
|
|
|
|
|
No education
|
|
|
|
.219[.081,.357] ***
|
|
Mother’s religion (Orthodox)
|
|
|
|
|
|
Muslim
|
|
|
|
.120[-009, .248] *
|
|
Others
|
|
|
|
.005[-.140, .149]
|
|
Maternal age at child birth: (<19, ref)
|
|
|
|
|
|
>19
|
|
|
|
-.397[-.549, -.244] ***
|
|
Maternal employment: No (ref)
|
|
|
|
|
|
Yes
|
|
|
|
.162[.062, .262] ***
|
|
Survey period: 2000-2011(ref)
|
|
|
|
|
|
2016-2019
|
|
|
|
-.618[-.748, -488] ***
|
|
Constant
|
-2.89[-2.98, -2.79] ***
|
-3.28[-3.46,3.10] ***
|
-2.64[-2.87, -2.40] ***
|
.893[.511, 1.275] ***
|
|
Random effect
|
|
|
|
|
Community level variance
|
.051 [.023,.108]
|
.054[.026, .111]
|
.048[.022,.106]
|
.077[.041, .145]
|
|
Household level variance
|
1.24 [1.03,1.50]
|
1.23 [1.012,1.49]
|
1.12[.914,1.372]
|
1.02[.783, 1.335]
|
|
ICC community
|
0.011
|
0.012
|
0.011
|
0.017
|
|
MOR community
|
1.239
|
1.248
|
1.232
|
1.302
|
|
ICC household
|
0.271
|
0.268
|
0.251
|
0.233
|
|
MOR household
|
2.891
|
2.877
|
2.744
|
2.623
|
|
AIC (smaller is better)
|
20580.7
|
20527.8
|
20284.27
|
16522.83
|
|
BIC (smaller is better)
|
20606.12
|
20578.65
|
20385.97
|
16760.12
|
|
LR test (Chi2)
|
214.65***
|
58.90***
|
255.53***
|
3793.44***
|
|
*** p<.01, ** p<.05, * p<.1
Source: Authors analysis using EDHS 2000-2016
3.4. Multivariate Decomposition Analysis Results
Table 4 provides a detailed multivariable decomposition analysis of maternal education-based inequalities in under-five mortality spanning the years 2000 to 2016. The breakdown reveals that 31% of the observed inequalities can be explained by differences in characteristic effects, while the remaining 69% is attributed to the unexplained component. Notably, efforts aimed at minimizing the gap in the number of children ever born and enhancing partner's education status emerge as crucial contributors to reducing maternal education-based inequalities. Specifically, addressing the disparity in the number of children ever born could lead to a substantial 61.09% reduction, and improving partner’s education status could contribute to a 24% reduction, both statistically significant at p<0.001. On the other hand, the analysis underscores the potential of household size to widen the gap in maternal education-based inequalities by 46.93%, signaling the need for targeted interventions in this regard. These findings emphasize the nuanced nature of maternal education's impact on under-five mortality and underscore specific areas for targeted public health interventions to address and alleviate existing inequalities.
Table 4:: Multivariable Decomposition of maternal education in under-five mortality in Ethiopia, 2000-2016
Under-five mortality
|
Coef.
|
se
|
p-value
|
[95% CI]
|
Percent
|
Overall
|
|
Explained gap (E)
|
0.009
|
0.002
|
0.000
|
[.005, .013]
|
31.00
|
Unexplained gap (C)
|
0.012
|
0.004
|
0.000
|
[.012, .027]
|
69/00
|
Due to Difference in Characteristics (E)
|
|
Household size
|
-0.013
|
0.001
|
0.000
|
[-.015, -.012]
|
--46.93
|
Source of drinking water
|
0.002
|
0.001
|
0.053
|
[-.0001, .003]
|
5.65
|
Child sex
|
0.004
|
0.001
|
0.000
|
[.002, .006]
|
0.02
|
Child size at birth
|
-0.001
|
0.0003
|
0.029
|
[-.0011, -.0001]
|
-2.19
|
Duration of breastfeeding
|
-0.001
|
0.0001
|
0.000
|
[-.0007, -.0005]
|
-2.14
|
Total children ever born
|
0.017
|
0.001
|
0.000
|
[.015, .019]
|
61.09
|
Partner’s education
|
0.001
|
0.002
|
0.000
|
[.003, .0105]
|
24.28
|
Maternal age at child birth
|
-0.002
|
0.0003
|
0.000
|
[-.003, -.00017]
|
-7.88
|
Maternal employment
|
-0.0003
|
0.0001
|
0.067
|
[-.001, .000017]
|
-0.89
|
Due to Difference in Coefficients (C)
|
|
Household size
|
0.018
|
0.005
|
0.002
|
[.007, .030]
|
64.58
|
Source of drinking water
|
0.012
|
0.004
|
0.001
|
[.005, .019]
|
42.15
|
Child sex
|
-0.005
|
0.003
|
0.117
|
[-.012, .001]
|
-18.75
|
Child size at birth
|
-0.003
|
0.002
|
0.083
|
[-.007, .004]
|
-11.14
|
Duration of breastfeeding
|
-0.001
|
0.003
|
0.859
|
[-.007, .006]
|
-1.99
|
Total children ever born
|
-0.001
|
0.002
|
0.589
|
[-.005, .003]
|
-4.10
|
Partner’s education
|
-0.0001
|
0.002
|
0.944
|
[-.003, .003]
|
-0.39
|
Maternal age at child birth
|
-0.001
|
0.001
|
0.662
|
[-.003, .002]
|
-2.04
|
Maternal employment
|
0.0002
|
0.003
|
0.948
|
[-.006, .007]
|
0.75
|
_cons
|
-0.0002
|
0.012
|
0.999
|
[-.023, .023]
|
-0.06
|
Source: Authors analysis using EDHS 2000-2016.
3.5. Inequalities in U5M using Concentration Index
Figure 1 visually depicts the concentration curve for under-five mortality based on maternal education. Positioned above the equality line, the concentration curve indicates a higher concentration of under-five mortality among children born to mothers with lower educational attainment in Ethiopia throughout the study period. The concentration index, quantified as -0.072 (p<0.001), confirms the presence of inequalities in under-five mortality. This negative value signifies that the inequality disproportionately affects children born to uneducated mothers in the country, further emphasizing the need for targeted interventions to address and rectify these disparities (see Figure 1).
3.6. Trend Analysis Results
Table 5 presents the results of the trend analysis of maternal education-based inequalities in under-five mortality, alongside other influential factors such as partner’s education, household size, and number of children ever born. These factors were selected based on the observed behavioral effects from the multivariate decomposition analysis model mentioned earlier. The decomposition rate analysis indicates a noteworthy reduction in overall absolute inequalities in under-five mortality by 0.14 during the period 2000-2005 and a more substantial decrease of 0.32 from 2000 to 2016, reflecting a notable 56.25% change.
Specifically, the analysis attributes approximately 43.12% of the decline in under-five mortality to the narrowed gap in maternal education during the survey period. Moreover, parental education, represented by partner’s education, contributes significantly, accounting for 56.95% of the observed decline. It is important to note, however, that household size emerges as a potential contributor to the widening gap in under-five mortality. These findings underscore the dynamic nature of maternal education-based inequalities and the multifaceted influences that various factors exert on the observed trends in under-five mortality during the specified time frames (see Table 5).
Table 5: Trends in under-five mortality inequality by maternal education in Ethiopia, 2000-2019.
Explanatory
variables
|
Survey periods
|
2000-2005
|
2000-2011
|
2000-2016
|
D*
|
%
|
D
|
%
|
D
|
%
|
Maternal education
|
-0.04
|
29.84
|
-0.11
|
37.74
|
-0.14
|
43.12
|
Partner’s education
|
-0.08
|
58.58
|
-0.16
|
52.99
|
-0.18
|
56.95
|
Household size
|
0.01
|
-7.17
|
0.008
|
-2.75
|
0.013
|
-3.93
|
Total children ever born
|
-0.03
|
18.75
|
-0.04
|
12.05
|
-0.01
|
3.86
|
Overall
|
-0.14
|
100
|
-0.30
|
100
|
-0.32
|
100
|
*Absolute difference in proportion of under-five mortality
Source: Authors construction using EDHS 2000-2019