Introduction to mpindex

This guide presents an illustration of a simple application of mpindex package for computing Multidimensional Poverty Index (MPI) using the Alkire-Foster (AF) counting method developed by Sabina Alkire and James Foster.

1. Installation

To install the mpindex package from CRAN:

install.packages('mpindex')

If you want to get the latest development version of mpindex, install it from GitHub. Note that you may need to install devtools.

# install.packages("devtools")
devtools::install_github('yng-me/mpindex')

Load the package once you have successfully completed the installation.

library(mpindex)

2. MPI specifications

The initial step is to prepare an MPI specification file which will serve as references in the computation as well as generation of output in the later part of the process. This file should contain information about MPI dimensions, indicators and their corresponding weights.

This file should also be easy to create using the most common and accessible file types such as .xlsx (Excel), .json, .csv, or .txt (TSV).

Built-in specification files

For convenience, mpindex has included built-in specification files (in different formats). Each file contains dimensions, indicators, weight, and other relevant information of the Global MPI.

To see the list of files available:

system.file("extdata", package = "mpindex") |> list.files()
#> [1] "global-mpi-specs.csv"  "global-mpi-specs.json" "global-mpi-specs.txt" 
#> [4] "global-mpi-specs.xlsx"

To use a built-in specification file, say the .csv file, use below script to first get the full path of the file.

specs_file <- system.file("extdata", "global-mpi-specs.csv", package = "mpindex")

Dimension	Indicator	Variable	Weight	Description
Global MPI – Dimensions, Indicators, Deprivation Cutoffs, and Weights
Health	Nutrition	nutrition	0.167	Any person under 70 years of age for whom there is nutritional information is undernourished.
Health	Child mortality	child_mortality	0.167	A child under 18 has died in the household in the five-year period preceding the survey.
Education	Years of schooling	year_schooling	0.167	No eligible household member has completed six years of schooling.
Education	School attendance	school_attendance	0.167	Any school-aged child is not attending school up to the age at which he/she would complete class 8.
Living Standards	Cooking fuel	cooking_fuel	0.056	A household cooks using solid fuel, such as dung, agricultural crop, shrubs, wood, charcoal, or coal.
Living Standards	Sanitation	sanitation	0.056	The household has unimproved or no sanitation facility or it is improved but shared with other households.
Living Standards	Drinking water	drinking_water	0.056	The household's source of drinking water is not safe or safe drinking water is a 30-minute or longer walk from home, roundtrip.
Living Standards	Electricity	electricity	0.056	The household has no electricity.
Living Standards	Housing	housing	0.056	The household has inadequate housing materials in any of the three components: floor, roof, or walls.
Living Standards	Assets	assets	0.056	The household does not own more than one of these assets: radio, TV, telephone, computer, animal cart, bicycle, motorbike, or refrigerator, and does not own a car or truck.
Source: Alkire, S., Kanagaratnam, U. and Suppa, N. (2020). ‘The global Multidimensional Poverty Index (MPI): 2020 revision’, OPHI MPI Methodological Note 49, Oxford Poverty and Human Development Initiative, University of Oxford.

User-defined specification file

You can also define and create your own specification file if you prefer to or if you happen to use different sets of dimensions and indicators. At the minimum, however, this file should contain the following columns/variables:

Dimension,
Indicator,
Variable,
Weight,
Description (optional).

Note that the order in which you put these columns does not matter and also the names are not case sensitive, but make sure to spell the column names correctly.

You may download the template here if you do not want to start from scratch: MPI specification file sample.

Using `define_mpi_specs`

Once you have prepared the specification file, load it using the define_mpi_specs function (type ?define_mpi_specs for a help text).

For demonstration purposes, we will use the built-in specification file, discussed above.

specs_file <- system.file("extdata", "global-mpi-specs.csv", package = "mpindex")
define_mpi_specs(specs_file)
#> # A tibble: 10 × 7
#>    dimension        indicator    variable weight description variable_name label
#>    <chr>            <chr>        <chr>     <dbl> <chr>       <chr>         <chr>
#>  1 Health           Nutrition    nutriti… 0.167  Any person… d01_i01_nutr… Heal…
#>  2 Health           Child morta… child_m… 0.167  A child un… d01_i02_chil… Heal…
#>  3 Education        Years of sc… year_sc… 0.167  No eligibl… d02_i01_year… Educ…
#>  4 Education        School atte… school_… 0.167  Any school… d02_i02_scho… Educ…
#>  5 Living Standards Cooking fuel cooking… 0.0556 A househol… d03_i01_cook… Livi…
#>  6 Living Standards Sanitation   sanitat… 0.0556 The househ… d03_i02_sani… Livi…
#>  7 Living Standards Drinking wa… drinkin… 0.0556 The househ… d03_i03_drin… Livi…
#>  8 Living Standards Electricity  electri… 0.0556 The househ… d03_i04_elec… Livi…
#>  9 Living Standards Housing      housing  0.0556 The househ… d03_i05_hous… Livi…
#> 10 Living Standards Assets       assets   0.0556 The househ… d03_i06_asse… Livi…

define_mpi_specs requires a specification file as its first argument. The default poverty cutoff is set to 1/3 (based on Global MPI). You can also define a list of poverty cutoffs by specifying in the .poverty_cutoffs argument to achieve gradient MPIs.

define_mpi_specs(
  .mpi_specs_file = specs_file, 
  .poverty_cutoffs = c(1/3, 0.2, 0.8)
)

If your dataset contains unique ID, like uuid, it is recommended to define it here using the .uid argument.

define_mpi_specs(
  .mpi_specs_file = specs_file, 
  .uid = 'uuid'
)

You can also set the aggregation level. Make sure it corresponds to the column name present in your dataset (see ?df_househod and more below).

define_mpi_specs(
  .mpi_specs_file = specs_file, 
  .poverty_cutoffs = c(1/3, 0.2, 0.8),
  .uid = 'uuid',
  .aggregation = 'class'
)

.unit_of_analysis, .source_of_data, and .names_separator are merely used for auto labels when generating the output later.

Note: define_mpi_specs returns a data frame of MPI specs defined in the specs file. By default, it saves a global option named mpi_specs which can be accessed using getOption('mpi_specs').

For our demonstration, we will use the use_global_mpi_specs() wrapper function, which yields the same result as calling define_mpi_specs and passing the built-in specs file.

use_global_mpi_specs(
  .uid = 'uuid',
  .aggregation = 'class'
)

3. Data preparation

The user of mpindex is assumed to have basic familiarity with the concept of tidy data as well as able to perform data wrangling and transformation using the tidyverse ecosystem. Under the hood, mpindex uses dplyr verbs to perform data manipulation.

We also assume that your dataset is already tidy and ready for analysis. See R for Data Science by Hadley Wickham and Garrett Grolemund if you need a refresher.

Dataset

For this demonstration, we will use two (2) synthetic datasets available within the package:

df_household household-level data (type ?df_household for more info)
df_household_roster individual-level data (type ?df_household_roster for more info)

library(dplyr)

glimpse(df_household)
#> Rows: 198
#> Columns: 21
#> $ uuid               <chr> "5dbec60a-ebda-47bd-ae18-3b017a221125", "8b70c208-8…
#> $ class              <chr> "Rural", "Rural", "Rural", "Rural", "Rural", "Rural…
#> $ drinking_water     <int> 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, …
#> $ toilet             <int> 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 2, 1, 1, 4, 1, 1, 1, …
#> $ with_child_died    <int> 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 1, 2, 2, 2, …
#> $ roof               <int> 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, …
#> $ walls              <int> 1, 1, 3, 3, 3, 5, 5, 1, 3, 1, 1, 4, 8, 2, 2, 10, 2,…
#> $ floor              <int> 1, 1, 1, 1, 1, 4, 1, 1, 1, 1, 1, 1, 5, 2, 4, 1, 2, …
#> $ electricity        <int> 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 2, 1, 1, 2, …
#> $ cooking_fuel       <int> 3, 5, 5, 5, 5, 5, 5, 3, 3, 1, 3, 5, 5, 5, 5, 5, 5, …
#> $ asset_refrigerator <int> 1, 1, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, …
#> $ asset_radio        <int> 0, 0, 1, 0, 0, 0, 0, 0, 1, 0, 1, 0, 0, 0, 0, 0, 1, …
#> $ asset_tv           <int> 1, 0, 0, 1, 1, 1, 1, 2, 1, 1, 1, 0, 0, 0, 0, 0, 0, …
#> $ asset_telephone    <int> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, …
#> $ asset_mobile_phone <int> 4, 3, 2, 2, 5, 2, 1, 6, 2, 4, 1, 2, 3, 0, 0, 2, 0, …
#> $ asset_animal_cart  <int> 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, …
#> $ asset_computer     <int> 0, 0, 0, 0, 1, 0, 0, 2, 0, 0, 0, 0, 0, 0, 0, 0, 0, …
#> $ asset_motorcycle   <int> 0, 0, 0, 1, 1, 0, 0, 1, 0, 0, 0, 1, 1, 0, 0, 1, 0, …
#> $ asset_bicycle      <int> 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 1, 0, 0, 0, …
#> $ asset_car          <int> 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, …
#> $ asset_truck        <int> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, …

glimpse(df_household_roster)
#> Rows: 905
#> Columns: 8
#> $ uuid                     <chr> "5dbec60a-ebda-47bd-ae18-3b017a221125", "5dbe…
#> $ line_number              <int> 1, 2, 3, 4, 1, 2, 3, 4, 1, 2, 1, 2, 3, 4, 1, …
#> $ class                    <chr> "Rural", "Rural", "Rural", "Rural", "Rural", …
#> $ sex                      <chr> "Male", "Female", "Female", "Male", "Female",…
#> $ age                      <int> 55, 48, 17, 10, 30, 26, 3, 19, 66, 62, 34, 34…
#> $ attending_school         <int> NA, NA, 1, 1, NA, NA, NA, 1, NA, NA, NA, NA, …
#> $ undernourished           <int> 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, …
#> $ completed_6yrs_schooling <int> 1, 1, 1, 2, 1, 1, NA, 2, 1, 1, 2, 1, 2, NA, 2…

4. Creating deprivation profile

Using `define_deprivation`

First, we need to create an empty list, and name it deprivation_profile (but feel free to name it whatever you like).

deprivation_profile <- list()

To create a deprivation profile for each indicator, we use the define_deprivation function (see ?define_deprivation) and add to the deprivation_profile list we created above. Make sure that the deprivation profile for each indicator matches the variable name declared in the specification file.

1. Heath dimension

1.1. Nutrition

For this indicator, we use the df_household_roster dataset. By default, define_deprivation sets the .collapse = FALSE. Since we need to collapse it to the household level, we need to set .collapse = TRUE.

deprivation_profile$nutrition <- df_household_roster |> 
  define_deprivation(
    .indicator = nutrition,
    .cutoff = undernourished == 1 & age < 70,
    .collapse = TRUE
  )

1.2. Child mortality

For child mortality, we use the df_household dataset. But unlike in nutrition, we do not need to provide the .collapse argument since it is not applicable here.

deprivation_profile$child_mortality <- df_household |> 
  define_deprivation(
    .indicator = child_mortality,
    .cutoff = with_child_died == 1
  )

2. Education dimension

2.1. Years of schooling

deprivation_profile$year_schooling <- df_household_roster |> 
  define_deprivation(
    .indicator = year_schooling,
    .cutoff = completed_6yrs_schooling == 2,
    .collapse = TRUE
  )

2.2. School attendance

deprivation_profile$school_attendance <- df_household_roster |> 
  define_deprivation(
    .indicator = school_attendance,
    .cutoff = attending_school == 2 & age %in% c(5:24),
    .collapse = TRUE
  )

3. Living standards dimension

3.1. Cooking fuel

deprivation_profile$cooking_fuel <- df_household |> 
  define_deprivation(
    .indicator = cooking_fuel,
    .cutoff = cooking_fuel %in% c(4:6, 9)
  )

3.2. Sanitation

deprivation_profile$sanitation <- df_household |> 
  define_deprivation(
    .indicator = sanitation,
    .cutoff = toilet > 1
  )

3.3. Drinking water

deprivation_profile$drinking_water <- df_household |> 
  define_deprivation(
    .indicator = drinking_water,
    .cutoff = drinking_water == 2
  )

3.4. Electricity

deprivation_profile$electricity <- df_household |> 
  define_deprivation(
    .indicator = electricity,
    .cutoff = electricity == 2
  )

3.5. Housing

deprivation_profile$housing <- df_household |> 
  define_deprivation(
    .indicator = housing,
    .cutoff = roof %in% c(5, 7, 9) | walls %in% c(5, 8, 9, 99) == 2 | floor %in% c(5, 6, 9)
  )

3.6. Assets

For this indicator, we need additional transformation.

deprivation_profile$assets <- df_household |> 
  mutate_at(vars(starts_with('asset_')), ~ if_else(. > 0, 1L, 0L)) |> 
  mutate(
    asset_phone = if_else(
      (asset_telephone + asset_mobile_phone) > 0, 
      1L, 
      0L
    )
  ) |> 
  mutate(
    with_hh_conveniences = (
      asset_tv + asset_phone + asset_computer + 
        asset_animal_cart + asset_bicycle + 
        asset_motorcycle + asset_refrigerator) > 1,
    with_mobility_assets = (asset_car + asset_truck) > 0
  ) |> 
  define_deprivation(
    .indicator = assets,
    .cutoff = !(with_hh_conveniences & with_mobility_assets)
  )

5. Computing the MPI

Using `compute_mpi`

After completing the deprivation profile, use the compute_mpi function and pass the deprivation_profile list as the first argument.

mpi_result <- df_household |>
  compute_mpi(deprivation_profile)

names(mpi_result)
#> [1] "index"              "contribution"       "headcount_ratio"   
#> [4] "deprivation_matrix"

Outputs

1. The MPI

mpi_result$index

Class	Number of cases	Headcount Ratio (H)	Intensity of Deprivation Among the Poor (A)	MPI (H x A)
MPI Results using 33% Poverty Cutoff
Rural	98	0.480	0.513	0.246
Urban	100	0.280	0.405	0.113

2. Contribution by dimension

mpi_result$contribution

Class	Number of cases	Health		Education		Living Standards
Contribution by Dimenstion and Indicator to MPI using 33% Poverty Cutoff
Class	Number of cases	Nutrition	Child mortality	Years of schooling	School attendance	Cooking fuel	Sanitation	Drinking water	Electricity	Housing	Assets
Rural	98	20.7	1.4	32.5	18.0	9.0	1.4	0.7	2.1	3.7	10.6
Urban	100	22.1	1.5	36.8	25.0	0.0	2.5	0.5	0.0	0.5	11.3

3. Headcount ratio

mpi_result$headcount_ratio$uncensored

Class	Number of cases	Health		Education		Living Standards
Uncensored Headcount Ratio
Class	Number of cases	Nutrition	Child mortality	Years of schooling	School attendance	Cooking fuel	Sanitation	Drinking water	Electricity	Housing	Assets
Rural	98	0.306	0.020	0.888	0.276	0.582	0.082	0.031	0.102	0.163	0.867
Urban	100	0.170	0.010	0.740	0.240	0.010	0.100	0.010	0.000	0.010	0.660

mpi_result$headcount_ratio$censored

Class	Number of cases	Health		Education		Living Standards
Censored Headcount Ratio using 33% Poverty Cutoff
Class	Number of cases	Nutrition	Child mortality	Years of schooling	School attendance	Cooking fuel	Sanitation	Drinking water	Electricity	Housing	Assets
Rural	98	0.306	0.020	0.480	0.265	0.398	0.061	0.031	0.092	0.163	0.469
Urban	100	0.150	0.010	0.250	0.170	0.000	0.050	0.010	0.000	0.010	0.230

4. Deprivation matrix (first 6 observations)

mpi_result$deprivation_matrix$uncensored |> head()

UUID	Class	Deprivation score	Health		Education		Living Standards
Uncensored Deprivation Matrix
UUID	Class	Deprivation score	Nutrition	Child mortality	Years of schooling	School attendance	Cooking fuel	Sanitation	Drinking water	Electricity	Housing	Assets
5dbec60a-ebda-47bd-ae18-3b017a221125	Rural	0.222	0	0	1	0	0	0	0	0	0	1
8b70c208-8642-408c-8a51-30bcaa106069	Rural	0.278	0	0	1	0	1	0	0	0	0	1
aa7cb64d-ba16-4842-8994-8772060f432d	Rural	0.111	0	0	0	0	1	0	0	0	0	1
df3e5c9b-7218-451d-9917-cd552cf4f40f	Rural	0.278	0	0	1	0	1	0	0	0	0	1
57babe6a-c163-4d8e-aa80-3a9bc9290d2d	Rural	0.278	0	0	1	0	1	0	0	0	0	1
ba3f75cd-102d-482d-a979-b9098e9e0e18	Rural	0.278	0	0	1	0	1	0	0	0	0	1

mpi_result$deprivation_matrix$censored |> head()

UUID	Class	Deprivation score	Health		Education		Living Standards
Censored Deprivation Matrix using 33% Poverty Cutoff
UUID	Class	Deprivation score	Nutrition	Child mortality	Years of schooling	School attendance	Cooking fuel	Sanitation	Drinking water	Electricity	Housing	Assets
5dbec60a-ebda-47bd-ae18-3b017a221125	Rural	0.000	0	0	0	0	0	0	0	0	0	0
8b70c208-8642-408c-8a51-30bcaa106069	Rural	0.000	0	0	0	0	0	0	0	0	0	0
aa7cb64d-ba16-4842-8994-8772060f432d	Rural	0.000	0	0	0	0	0	0	0	0	0	0
df3e5c9b-7218-451d-9917-cd552cf4f40f	Rural	0.000	0	0	0	0	0	0	0	0	0	0
57babe6a-c163-4d8e-aa80-3a9bc9290d2d	Rural	0.000	0	0	0	0	0	0	0	0	0	0
ba3f75cd-102d-482d-a979-b9098e9e0e18	Rural	0.000	0	0	0	0	0	0	0	0	0	0

Note: Deprivation matrices are included by default when you run compute_mpi. If you want to exclude these in your output, set .include_deprivation_matrix equal to FALSE.

6. Saving output

You may also save your output into an Excel file. You may choose to format the output or retain its tidy format by setting the formatted_output parameter appropriately.

# Formatted output
save_mpi(mpi_result, .filename = 'MPI Sample Output')

# Not formatted
save_mpi(mpi_result, .filename = 'MPI Sample Output (no format)', .formatted_output = FALSE)

Full script

# ----------------------------------
# Load MPI specs from the built-in specs file 
use_global_mpi_specs(
  .uid = 'uuid',
  .aggregation = 'class'
)

# ----------------------------------
# Create an empty list to store deprivation profile for each indicator
deprivation_profile <- list()

deprivation_profile$nutrition <- df_household_roster |>
  define_deprivation(
   .indicator = nutrition,
   .cutoff = undernourished == 1 & age < 70,
   .collapse = TRUE
  )

deprivation_profile$child_mortality <- df_household |>
  define_deprivation(
   .indicator = child_mortality,
   .cutoff = with_child_died == 1
  )

deprivation_profile$year_schooling <- df_household_roster |>
  define_deprivation(
   .indicator = year_schooling,
   .cutoff = completed_6yrs_schooling == 2,
   .collapse = TRUE
  )

deprivation_profile$school_attendance <- df_household_roster |>
  define_deprivation(
   .indicator = school_attendance,
   .cutoff = attending_school == 2 & age %in% c(5:24),
   .collapse = TRUE
  )

deprivation_profile$cooking_fuel <- df_household |>
  define_deprivation(
   .indicator = cooking_fuel,
   .cutoff = cooking_fuel %in% c(4:6, 9)
  )

deprivation_profile$sanitation <- df_household |>
  define_deprivation(
   .indicator = sanitation,
   .cutoff = toilet > 1
  )

deprivation_profile$drinking_water <- df_household |>
  define_deprivation(
   .indicator = drinking_water,
   .cutoff = drinking_water == 2
  )

deprivation_profile$electricity <- df_household |>
  define_deprivation(
   .indicator = electricity,
   .cutoff = electricity == 2
  )

deprivation_profile$housing <- df_household |>
  define_deprivation(
   .indicator = housing,
   .cutoff = roof %in% c(5, 7, 9) | 
     walls %in% c(5, 8, 9, 99) == 2 | 
     floor %in% c(5, 6, 9)
  )

deprivation_profile$assets <- df_household |>
  dplyr::mutate_at(
    dplyr::vars(dplyr::starts_with('asset_')), 
    ~ dplyr::if_else(. > 0, 1L, 0L)
  ) |>
  dplyr::mutate(
   asset_phone = dplyr::if_else(
     (asset_telephone + asset_mobile_phone) > 0,
     1L,
     0L
   )
  ) |>
  dplyr::mutate(
   with_hh_conveniences = (
     asset_tv + asset_phone + asset_computer +
       asset_animal_cart + asset_bicycle +
       asset_motorcycle + asset_refrigerator) > 1,
   with_mobility_assets = (asset_car + asset_truck) > 0
  ) |>
  define_deprivation(
   .indicator = assets,
   .cutoff = !(with_hh_conveniences & with_mobility_assets)
  )

# ----------------------------------
# Compute the MPI
mpi_result <- df_household |>
  compute_mpi(deprivation_profile)

# ----------------------------------
# You may also save your output into an Excel file

# Formatted output
# save_mpi(mpi_result, .filename = 'MPI Sample Output', .include_specs = T)

# Not formatted
save_mpi(
  mpi_result, 
  .filename = 'MPI Sample Output (no format)', 
  .formatted_output = FALSE, 
  .include_specs = TRUE
)