# ! # # # ! # # # ! # # ! # #months old, are stunted, underweight, and wasted, Anthropometric Data- "#$%#&!'()#$*(+$*!,+! #respectively. Additionally, note the difference in S.+9."F#6+7)/3#-";#6)-)26#/4#9/3<#6*/96#)*-)# prevalence estimates before and after outlier trimming; Comparing ERSS and DHS -(.#/(0++%*!($!1(2#&(3! E+"# stunting-3+# E/3+# drops from0.<+0@# 52 to 48 )/#percent 8+# 4-3E."F# after trimming )*-"# the 9/E+"(#9*/#-3+#E/3+#0.<+0@#)/#8+#9/3<."F#."# z-scores. In!"#$%&%'$%&&(#)*+#,-)./"-0#123+-2#/4#5)-).6).76#."# 2011/2012, Ethiopiaʼs Central Statistical Agency, in -# */26+*/0;# +")+3I3.6+G#M/3+# )*-"# *-04#/4#-00# 7/00-8/3-)./"#9.)*#)*+#:/30;#1-"<#7/";27)+;#)*+# collaboration with the World Bank, conducted the first 9/3<."F# 9/E+"# -3+# +"F-F+;# -# "/"C4-3E# ."# trimming Table 1: Rural ERSS, pre- and post- =+"+3-0# wave >/26+*/0;# of the 523?+@# Ethiopia Rural A-"+0# B=>5CA-"+0D(# Socioeconomic Survey */26+*/0;# 826."+66(# 9*.7*# .6# 262-00@# ?+3@#   Prevalence  (Std.  Error)   9*.7*# 7/00+7)6# ;+)-.0+;# ;-)-# /"# ."7/E+# 9.)*# -# 6E-00C67-0+#-";#."4/3E-0G#:-F+#9/3<#4/3#8/)*# (ERSS), which collects detailed data on household Pre-­‐trimming   Post-­‐trimming   4/726# /"# F+";+3G# H*+# =>5CA-"+0# .6# -# "-)./"-00@# F3/2I6# .6# I3+;/E."-")0@# /2)6.;+# -F3.720)23+(# welfare and income-generating activity. The ERSSʼ Stunted   52%          (.02)   48%          (.02)  -6# 3+I3+6+")-).?+# 623?+@# 9.)*# -# 6-EI0+# /4# J(%%%# -";# E+"# -3+# E/3+# )*-"# )9.7+# -6# 0.<+0@# sample includes */26+*/0;6K# )*+#4,0006+7/";#households 9-?+# 32"6# that 43/E# are 9/E+"# )/# *-?+# Underweight   9-F+# +EI0/@E+")# 27%          (.02)   27%          (9*+"# .02)   representative $%&$'$%&LG# H*.6#of small towns "/)+# and rural I3+6+")6# areas; 6)-).6).76# 6+0+7)+;# wave 2 9/3<."FG## will be expanded to include urban areas and will run from /"#F+";+3#-";#0.?+0.*//;6#."#,.F+3.-G## Wasted   11%          (.01)   11%          (.01)   # 2013 to 2014. This note evaluates the performance of !"#$%&"'()*+),)-.*,(/*+'0)%1*%2*3%14** !"#$%&"'()*+',)%-*.&*/'(0'- the ERSS-wave 1 anthropometric indicators through ! a Comparison ;5 ;5?( with DHS series M+"# of checks and data9/E+"# -";# ."# comparisons. )*+# =>5CA-"+0# 3+I/3)# )*+.3# :5 The :5?( Demographic Health Survey (DHS) focuses on E-."# +EI0/@E+")# 6+7)/3(# 9*.7*# 7/20;# 8+# 9-F+# 95 95?( population and health and follows a rigorous ERSS and Child Anthropometric 9/3<(#*/26+*/0;#+")+3I3.6+#-7).?.).+6(#/3#4-3E."FG# 85 85?( 75 75?( methodology for anthropometric data collection. Drawing NF3.720)23+(# )3-;+(# -";# I+36/"-0# 6+3?.7+6# -3+# Indicators .EI/3)-")#+EI0/@E+")#6+7)/36#4/3#8/)*#E+"#-";# 65 65?( comparisons5 5?( between 2011 DHS data and recent ERSS <,& 9/E+"G#:*.0+#-F3.720)23+#.6#)*+#0-3F+6)#6+7)/3#4/3# The three anthropometric indicators most often results can help identify potential gaps =%>,& in ERSSʼ E+"(# .)# .6# referenced for )3-;."F# monitoring )*-)# ;/E."-)+6# malnutrition 4/3# 9/E+"# in children are: methodology as well as help data users assess data BLOPD# 9.)*# -F3.720)23+# ."# 6+7/";# I0-7+# stunting, or low height-for-age; underweight, low BLLPDG# weight-!"# quality. One notable difference between the ERSS and 7/EI-3.6/"(# E+"# -3+# E27*# 0+66# for-age; and wasting, low weight-for-height. More 0.<+0@# )/# 8+# DHS is the criterion for age eligibility; Ethiopiaʼs DHS )3-;."F#B&%PD#)*-"#."#/)*+3#6+7)/36G#:/E+"#9*/# collects weight and height measurements for all children specifically, these figures represent children whose * 9/3<#-3+#-06/#E/3+#0.<+0@#)/#8+#."#E-"24-7)23."F# less than 5 years old (including * 0-6 months). To height-for-age, weight-for-age, and weight-for-height fall )*-"#-3+#E+"#9*/#9/36; weight-for-age z-scores (WAZ) <-6 or   Prevalence  (Std.  Error)   Other Sectors 8.2 4.0 # >5; # and weight-for-height (WHZ) z-scores <-5 or >5. 3 W+E-0+# I-3).7.I-)./"# ."# 6+04C+EI0/@+;# ERSS   DHS   /44C ,/)+Q#R)*+3#5+7)/36#."702;+6#E."."F(#*+-0)*(# Approximately 300 children (13 percent) had at least one Stunted   +X7++;6# )*-)# 4-3E# -7).?.).+6# 48   /4#          (.02)   E+"# ."# 8/)*# 50            (.01)   4."-"7.-0'."623-"7+'3+-0#+6)-)+#6+3?.7+6(#+0+7)3.7.)@'# Underweight***   27            (.02)   )*+#"/3)*#-";#)*+#6/2)*G#!"#)*+#"/3)*(#4+E-0+# 33            (.02)   malnutrition estimate excluded from the analysis based 9-)+3'F-6'9-6)+(#I3/4+66./"-0'67.+").4.7')+7*".7-0# on these trimming rules. I-3).7.I-)./"#Wasted  .6# JJ# I+37+")(# 11            (.01)   7/EI-3+;#        (.01)   10    $&# )/# -7).?.).+6(#-";#/)*+3G# ***Difference  between  means  is  statistically  significant  at  p<0.01   Table 1 shows the stunting, underweight, and wasting, !"#$%&"'()*+,"*(%-(.#"#$/#$0/()+$,-((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((1112&$3,+$"&/#"#23%42&3( prevalence ( estimates for rural areas. In rural Ethiopia, We find that the medians for all three indicators are 48 percent, 27 percent, and 11 percent of children 6-59 statistically different between surveys. While stunting prevalence estimates are similar between ERSS and 7556 +#"+35' ?56 !%(1%+@ >56 >5 >56   =56 =5 =56( <56 A3+$0*'#*+, <5 <56( ;56 DHS, the left-hand tail of the HAZ distribution (those with prevalence points for Tigray and Oromia A3+$0*'#*+, that are five :56 ;5 ;56( height-for-age z-scores more than 4 deviations from the percentage :5 :56( points higher than ERSS ʼ respective 956 B%*/,C%'D median) is much thicker 856 for the ERSS sample (see estimates. 95 956( Chi-squared tests of independence B%*/,C%'D confirm ,&#,+E+$/, Figure 1). Post-collection 756 analysis of these cases that the85 regional 856( distributions for ERSS ,&#,+E+$/, 1are and DHS revealed a combination 56 of data measurement andF"3,(1%+@entry similar for75 756( stunting and wasting prevalence estimates . error for height values. Comparing the two $&("3 WAZ 5 56( F"3,(1%+@ distributions can help shed light on the difference F"3,(1%+@ in Table 3: ERSS H"', vs. I,J"', DHS, by H"', region I,J"', underweight prevalence estimates of 6 percentage $&(&%&G"3   B,"D B,"D B,"D Stunted  (%)   B,"D Wasted  (%)   K+L"& K+L"& M*+"' M*+"' points; the ERSS distribution is centered a bit to the left ' (Std.  Error)   (Std.  Error)   ) of the DHS distribution (see Figure 2). As seen in Figure ) Region   ERSS   DHS   ERSS   DHS   3, the overall shapes of the WHZ distributions looks very Tigray   53  (.05)   58  (.02)   7  (.03)   11  (.01)   similar, though we 62"%'(+#*'+$&171&1"3'+#"'"8+*1%",')%'+'#"91)%+:' note the DHS distribution is much 5%-(%1)*-()3"14)67)816*-9:#1*') ! ;+313<' 2)="7"#<' &2"' #"34:&3' +9+1%' #"(:"$&' *+:"' Amhara   52  (.04)   55  (.02)   10  (.02)   11  (.01)   tighter (i.e., more concentrated M))G1%9' +&' 4#;+%' +%,' #4#+:' :)$+&1)%3' 3"!+#+&":E<' ,)*1%+%$"' at 1%'the median). =1&2' >?' !"#$"%&' Oromia   (+#*1%9' 43  (.03)   48  (.02)   10  (.02)   10  (.01)   #"34:&3' 32)=' +' 2192"#' !+#&1$1!+&1)%' 1%' =+9"' !+#&1$1!+&1)%' +*)%9' *"%' 7"#343' @A' !"#$"%&' SNNP   "*!:)E*"%&' 53  (1%' .04)   49   4#;+%' (.02)   #4#+:' 7"#343' .02)   14  (I19"#1+<' 8   =1&2' (.01)   Figure 1: Height-for-age +*)%9' z-score =)*"%' distributions 1%' &2"' %)#&25' B2"' ,1(("#"%$"3' 1%' All  =+9"'"*!:)E*"%&'1%'4#;+%'+#"+3'#"!)#&",'+&'CA' other   &2"'3)4&2'+#"'%)&'+3'9:+#1%9<'=1&2';)&2'*"%'+%,' HAZ distribution, ERSS HAZ distribution, DHS 39  (.04)   47  (.02)   10  (.02)   18  (.02)   regions   !"#$"%&' $)*!+#",' &)' @?' !"#$"%&' 1%' #4#+:' +#"+35' 15 15 =)*"%' #"!)#&1%9' +#)4%,' CA' !"#$"%&' N"9+#,:"33' )(' 9")9#+!21$+:' :)$+&1)%<' *"%' +#"' !+#&1$1!+&1)%'#+&"3'1%'+9#1$4:&4#"5' *)#"' :1G":E' &)' ;"' 1%7):7",' 1%' =+9"' "*!:)E*"%&' Summary 10 10 &2+%' =)*"%5' O#;+%' +#"+3' #"!)#&' *+:"' !"#$%&"'()!%*($&+,)*-().%/0"1) % % Relative to DHS, the !+#&1$1!+&1)%' ERSS +&' data appear /P' !"#$"%&' to underperform +%,' ("*+:"' +&' C@' 5 5 "2)3"14'' slightly in terms of outliers (2 percent for !"#$"%&5' N4#+:' +#"+3' #"!)#&' *+:"' !+#&1$1!+&1)%'DHS and +&' 13 D8+*1%1%9' !+#&1$1!+&1)%' +&' &2"' 2)43"2):,' percent for ERSS). However, this @-'!"#$"%&'+%,'("*+:"'+&'0'!"#$"%&5'' evidence is somewhat 0 0   +&' :"+3&' )%"' 2)43"2):,' *"*;"#' -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 Height-for-age z-score Height-for-age z-score :"7":' =2"#"' tempered by differences in data preparation processes. Figure 2: Weight-for-age !+#&1$1!+&"3' z-score 1%' &2"' +$&171&E<' #"34:&3' 32)=' &2+&' distributions fields can 1%' Q+#&1$1!+&1)%' DHS data be )((F(+#*' overwritten 3":(F"*!:)E*"%&' at the CSA offices 13' if *+:"F2"+,",' 2)43"2):,3' WAZ distribution, ERSS +#"' WAZ distribution, DHS *)#"' :1G":E' &)' values2192"#' are deemed+*)%9' =)*"%' &2+%' non-credible, *"%' while 1%' ;)&2' ERSS 4#;+%' data errors !+#&1$1!+&"' 1%' +::' +$&171&1"3' $)*!+#",' &)' +%,'in are fixed the +#"+35' #4#+:' R%' 4#;+%' +#"+3<' field. Comparisons )((F(+#*' of ERSS and3":(F DHS 15 15 ("*+:"F2"+,",' 2)43"2):,35' D%9+9"*"%&' 1%' "*!:)E",'!+#&1$1!+&1)%'+*)%9'=)*"%'3&+%,3'+&' anthropometric data reveal similar prevalence estimates, 2)43"2):,' "%&"#!#13"' 13' &2"' *)3&' $)**)%' 00'!"#$"%&'$)*!+#",'&)'-0'!"#$"%&'()#'*"%<'+%,' 10 10 z-score distributions, and regional patterns between the Mild("*+:"' 1%' #4#+:' +#"+3' )((F(+#*' 3":(F"*!:)E*"%&' % % +$&171&E' ()#' ;)&2' *+:"F' +%,' ("*+:"F2"+,",' two surveys. evidence indicates that, at the 2)43"2):,3H'0@'+%,'>-'!"#$"%&'#"3!"$&17":E5' 3&+%,3'+&'-/'!"#$"%&'$)*!+#",'&)'@P'!"#$"%&'()#' national level, there is a difference between the two 5 5 ' *"%5' S)="7"#<' underweight indicators,()#' (+#*1%9<' though &2"' the #"7"#3"' overall 13' &2"' WAZ 1%' 4#;+%' are $+3"H' patterns +#"+3<' *"%' +#"' !#",)*1%+%&' =1&2' 0 0 distribution comparable. Formally, we find   -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 -6 -5 -4 -3 -2 -1 0 1 2 3 Weight-for-age z-score =5 Weight-for-age z-score =56( ( significant &)'difference @C' !"#$"%&' $)*!+#",' no statistically in =)*"%5' .' !"#$"%&' ()#' regional <5 Figure 3: Weight-for-height <56( z-score distributions B213' 13' of distributions +:3)' &2"' $+3"' stunting and 1%' wasting; #4#+:' +#"+3<' =1&2' >.' however, the WHZ distribution, ERSS ( ;5 ;56( WHZ distribution, DHS !"#$"%&' )(' *"%' !+#&1$1!+&1%9' patterns may lend themselves to varying 1%' (+#*1%9' policy A3+$0*'#*+, $)*!+#",'&)'/-'!"#$"%&')('=)*"%5' 15 15 ( interpretations. :5 :56( (   B213' is;#1"(' 13' ;+3",' ,+&+' $)::"$&",' )%' collected ;E' &2"' 10 10 95 956( B%*/,C%'D This brief based on data by the Central % % ( ,&#,+E+$/, I+&1)%+:'T4#"+4')('U&+&13&1$3'+3'!+#&')('&2"'M171%9' Statistical Agency as part of the Living Standards 85 U&+%,+#,3' V"+34#"*"%&' – Integrated U&4,E' R%&"9#+&",' ²' on 856( 5 5 ( F"3,(1%+@ Measurement Study Surveys Agriculture 75 756( U4#7"E3' )%' W9#1$4:&4#"' JMUVUFRUWL' !#)X"$&5' B2"' (LSMS-ISA) project. The full dataset is available for (4::',+&+3"&'13'+7+1:+;:"'()#',)=%:)+,'+&'IWYW'71+' 0 0 (   -5 -4 -3 -2 -1 0 1 2 3 4 5 -5 -4 -3 -2 -1 0 1 2 3 4 5 download at CSA via http://www.csa.gov.et. Weight-for-height z-score Weight-for-height z-score 5 56( ===5%19"#1+%3&+&59)75%95'' A'' H"', I,J"', The two sets of data are comparableB,"D at B,"D the national The findings outlined in this brief are drawn from… level, but we find greater differences at the regional ' ' I%+(J%+,($&-%+J"#$%&N(E',"/,(4$/$#O( Revisiting z-scores: 1112&$3,+$"&/#"#23%42&3( A review of LSMS-ISA level. The ERSS was stratified regionally and has five Anthropometric ' Data , The World Bank, as presented at ( I)&' 34#!#131%9:E<' domains of analysis. These domains #4#+:' +%,' &2"' include the 4#;+%'four ,171,"' P+(0%&#"0#O( the LSMS-ISA Annual Workshop 2013, Addis Ababa, largest regions, 32)=3' *)#"' Tigray, +9#1$4:&4#+:' Amhara, Oromia,2)43"2):,3' and the 1%' &2"' #4#+:' !"#$%&"'()*+,"*(%-(.#"#$/#$0/( +#"+3' ()#' ;)&2' *+:"F' +%,' ("*+:"F' 2"+,",' Ethiopia -,,DL"0@Q&$3,+$"&/#"#23%42&3( Southern Nations, Nationalities, and Peoples (SNNP), as well as one domain representing all other regions. Table   3 provides the !"#$%&"'()*+,"*(%-(.#"#$/#$0/()+$,-((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((1112&$3,+$"&/#"#23%42&3( breakdown of stunting and wasting ( prevalence for these four regions, as calculated from ERSS and DHS data. While the ERSS data reports 14                                                                                                                 1 2 2 percent of children from SNNP are wasted, the DHS Chi-squared test of independence uses Χ =(O-E) /E, where O and E are the observed and expected number of observations, respectively. indicates only 8 percent of children meet the wasting “Expected” refers to the number of observations (from the ERSS criteria in the region. DHS also reports stunting sample) meeting some criteria one would expect to see, treating the DHS as reflective of the population distribution.