Summary of this file

I get IPC and FAO DIEM data and merge it by admin level 2 and year. Then I provide some desciptives of the food security indicators for the entire sample (of merged data) and summarize those indicators by IPC phase - to see how IPC matches to these food security indicators.

Data work

Here is a summary of the data work:

  • Get IPC phase per country per area (admin level 2)
  • Take the DIEM food security module data, which contains several food security indicators like food consumption scores, reduced coping index, etc. The income module looked less relevant, but at the link in the next bullet point you can take a look at other modules and variables available. To me, nothing from DIEM looks suitable for estimating calorie gaps.
  • Selected what indicators to include (see the list in the section below). To see all variables for all modules, see this file: https://www.dropbox.com/scl/fi/rsagth53r1q0pxdrpktxh/DIEM_2023_-_Fields_descriptions.xlsx?rlkey=vn9t7po3fwla9c1ahvz0hwfgj&dl=0
  • I merged DIEM data and IPC phase data by admin level 2 columns in both datasets and by time period of the indicator/IPC analysis. For the sake of speed, I didn’t touch any values in the admin level 2 columns of either dataset; I just merged what was already clearly a pair. For merging by time period, I created a window of time based on the start date of the survey/IPC analysis and then filtered as follows: start_dateForMatching_IPC <= end_dateForMatching_DIEM & end_dateForMatching_IPC >= start_dateForMatching_DIEM.


Selected indicators

Below are the key indicators, categorized by type. In the data shown below, I include even a more filtered list:

Food Insecurity Experience Scale (FIES)
- fies_rawscore_median, fies_rawscore_wmean: Median and weighted mean FIES raw scores.
- p_mod_median, p_mod_wmean: Prevalence of recent moderate or severe household food insecurity (FIES). Values range from 0 to 1.
- fies_rawscore_0 to fies_rawscore_8: Breakdown of FIES raw scores.

Livelihood Coping Strategy Index (LCSI)
- lcsi_0 to lcsi_3: Different levels of livelihood coping strategies.

Household Diet Diversity Score (HDDS)
- hdds_class_1 to hdds_class_3: Classification of household diet diversity.

Household Hunger Scale (HHS)
- hhs_0 to hhs_6: Measures of household hunger severity.

Food Consumption Score (FCS)
- fcs_median, fcs_wmean: Median and weighted mean Food Consumption Scores.

Food Consumption Groups (FCG)
- fcg_1 to fcg_3: Classification of households based on food consumption. Additional scores per food group are available if needed.

Reduced Coping Strategies Index (rCSI)
- rcsi_score_median, rcsi_score_wmean: Median and weighted mean reduced coping strategies index scores.





Viewing the data



Food consumption score

FCS distribution across all matched data

FCS median score at admin 2 level summarized by IPC phase

Food Consumption Scores summarized by IPC phase
area_overall_phase variable mean median sd min max count
1 fcs_median 47.4 50.0 10.0 27 66.5 70
2 fcs_median 45.1 48.5 13.1 20 71.0 170
3 fcs_median 38.6 36.0 13.2 12 79.0 379
4 fcs_median 34.4 28.0 21.3 12 79.0 9


FCS using the district level weighted mean

FCS w mean score at admin 2 level summarized by IPC phase

Food Consumption Scores (weighted mean) summarized by IPC phase
area_overall_phase variable mean median sd min max count
1 fcs_wmean 49.0 51.06695 8.9 26.82486 69.01293 70
2 fcs_wmean 45.4 47.53283 11.8 20.62376 69.40800 170
3 fcs_wmean 39.3 37.12467 11.9 13.30284 75.41667 379
4 fcs_wmean 36.2 31.12117 19.1 13.30284 75.41667 9



FCS gap calculation

WFP indicates a FCS >35 as adequate. Therefore, the deficit is calculated based on this.

\[\begin{equation} \text{Consumption Gap}_i = \begin{cases} \frac{z - y_i}{z}, & \text{if } y_i < z \\ 0, & \text{if } y_i \geq z \end{cases} \end{equation}\]

where \(y_i\) is the district-level median Food Consumption Score (FCS), and \(z\) is the FCS adequacy threshold of 35.

\[\begin{equation} \text{Food Consumption Gap Index by IPC Phase} = \frac{1}{N} \sum_{i=1}^{N} \text{Consumption Gap}_i \end{equation}\]

where \(i\) is the district and \(N\) is the number of districts.


cutoff <- 35
IPC_DIEM_withFCSCalculation <- IPC_DIEM %>%
  select(iso3, adm_name, DIEM_startDate,fcs_median, IPC_country_title: last_col() ) %>%
  mutate(FCS_povGapMeasure = (cutoff - fcs_median)/cutoff) %>%
  mutate(FCS_povGapMeasure = case_when(
    FCS_povGapMeasure < 0 ~ 0,
    TRUE ~ FCS_povGapMeasure) 
    ) %>%
  relocate(FCS_povGapMeasure, .after = fcs_median)
Average FCS gap (defined as below a score of 35) per IPC phase
area_overall_phase FCS_povGapMeasure_ByIPCPhase
1 0.02122449
2 0.05722689
3 0.10388240
4 0.24444444

Now we would need to associate these consumption gaps with calorie gaps. There is some work done on this like in Wiesmann et al. (2009). All related papers are in our resource folder in the shared folder: C:DropboxRice_IPC_analysis.

Here is one particularly relevant table from that paper:

Thresholds for creating calorie consumption groups

Calorie consumption in kilocalories/capita/day Shortfall, in percentage Profile
< 1,470 > 30 Poor
≥ 1,470 – < 2,100 ≤ 30 – > 0 Borderline
≥ 2,100 0 Acceptable

Source: Food consumption shortfalls described in World Food Programme (2005, 139).




Reduced coping strategies index

rcsi distribution across all matched data

RCSI median score at admin 2 level summarized by IPC phase

RCSI scores summarized by IPC phase
area_overall_phase variable mean median sd min max count
1 rcsi_score_median 5.8 4 5.6 0 21.5 70
2 rcsi_score_median 9.7 9 6.9 0 35.0 169
3 rcsi_score_median 12.4 11 7.4 0 46.0 377
4 rcsi_score_median 13.4 16 7.7 0 22.0 9

Fies_rawscore_median

FIES raw score - distribution across all matched data

FIES median raw score at admin 2 level summarized by IPC phase

FIES Raw Scores summarized by IPC phase
area_overall_phase variable mean median sd min max count
1 fies_rawscore_median 2.1 1 1.9 0 7 71
2 fies_rawscore_median 3.6 3 2.0 0 8 214
3 fies_rawscore_median 4.6 4 2.1 0 9 427
4 fies_rawscore_median 4.1 4 1.6 2 7 9