New package -mpitb- now available from SSC

[Nicolai Suppa]

Thanks to Kit Baum, the package mpitb is now available from SSC. Type

ssc install mpitb

to install the package.

The package requires Stata 16 or newer and is also available from Gitlab.

mpitb is a toolbox to estimate and analyze multidimensional poverty indices (MPI), which supports the popular Alkire-Foster framework to multidimensional poverty measurement.

The accompanying paper and slides provide further background information and examples.

If you run into any bugs, cryptic error messages or incomprehensible parts in the documentation, please report under the issue tracker.

Any kind of feedback is always welcome, whether reporting back how mpitb is useful for your work or where you'd like to see improvements. For extensions and feature requests, please also use the issue tracker.

I will announce major updates or revisions in this topic.

[Emmanuel Kisaame]

Hi Nicolai,

Thanks for the tool box. It's super useful. I'm evaluating the effect of a cash transfer program on child deprivation using both mpi and propensity score matching. However, when I run the mpitb est command, the rest of the variables (my covariates and treatment variable) disappear leaving only the mpi related variables. As a result, my matching command returns an error because the treatment variable cannot be found. Please advise on how I can go around this. Thank you so much.

Emmanuel

[Nicolai Suppa]

Hi Emmanuel,

Happy to hear that you find the toolbox useful! From your description it's not entirely clear to me how exactly you'd like to implement the propensity score matching (i.e. with which commands and which outcomes). Moreover, the toolbox is usually not touching or even dropping other variables in the dataset. Feel free to send me a personal message with further details, if that helps.

Nicolai

[Pablo Gonzalez]

Hi Nicolai,
Would it be feasible to estimate the index by region × sex ?

Pablo

[Nicolai Suppa]

Hi Pablo,

Two-way disaggregations are currently not supported out of the box. However, an easy workaround that often works is to generate a new variable reflecting both `region` and `sex` of an observation, for instance, with

egen regsex = group(region sex) , lab

You can use this variable for estimation purposes:

mpitb est ... , ... over(regsex) ...

To facilitate correct labels in the results file, you can save your labels from the microdata using

lab save regarea using lb_regarea , replace

After having created an appropriate variable in your results file, you can run this do-file to assign the correct value labels.

Important: Please consult the documentation of your data to ensure such an analysis can be done, which usually depends on the (complex) survey design.

Nicolai

[George Ford]

Your link to the paper is broken.

[Nicolai Suppa]

Hi George,

Thanks for reporting! Meanwhile, the paper was actually published in the Stata Journal.

Nicolai

[George Ford]

I've been working on something related and curious if I can use this. I'm reproducing a study that uses median income (categorized) as a measure of poverty (lowest income class = poor). There are many measures of poverty of course -- persons below x% poverty level, college education, broadband adoption, etc... The data is aggregated (zip or county). I used factor analysis to construct an index of poverty using many of these sorts of variables. Does this approach allow something similar? As it is aggregated, all the variables are continuous.

[Pablo Gonzalez]

Hi Nicolai,

thanks for your help.