dea and malmq: Data Envelopment Analysis and Malmquist Productivity Analysis

[William Lisowski]

The user-written dea command (for Data Envelopment Analysis) and malmq command (for Malmquist Productivity Analysis) can be dowloaded with supporting material and examples, along with an accompanying PowerPoint conference presentation, from the Download tab on the following page

https://ideas.repec.org/p/boc/chic11/21.html

from IDEAS at RePEc.

[William Lisowski]

At

https://www.statalist.org/forums/for...uctivity-index

Serena Chang advises that the malmq command requires a space on either side of the equal sign separating the input and output variable lists.

[William Lisowski]

At

https://www.statalist.org/forums/for...88#post1498788

I present the changes that need to be made to malmq.ado in order for the malmq command save option to not generate an error message about a missing file_exists() function.

[William Lisowski]

At

https://sourceforge.net/projects/deas/

you can find the Sourceforge site maintained by the author of the dea and malmq commands, if you are knowledgeable about retrieving software from Sourceforge (I am not).

[William Lisowski]

At the time of this post, it appears that the Sourceforge source given in post #4 is the most current version of the dea and malamq packages. The latest comment in dea.ado has a date of April 30 2014 and was released to Sourceforge on May 10 2014. The author has not been as good about commenting in malmq.ado, but that version was released on January 5 2013 and is clearly later than the similarly-numbered version available on in the other locations; specifically correcting the bug in the earlier version discussed in the link in post #3.

To download the current versions from Sourceforge, click on the link in post #4, then in the page that opens click on the blue "Browse Code" button, and in the page that then opens, click on the "trunk" folder. From the page that opens, you can download the dea and malmq .ado and .sthlp files.

To install these versions of dea or malmq, I recommend first doing in Stata

adopath
ado uninstall dea

to be sure that any previous version is uninstalled from your Stata PLUS directory, then copy the new versions into the PERSONAL directory indicated by the adopath command.

[William Lisowski]

The recommended un-installation code in post #5 is incomplete; it does not uninstall the dea package hosted at Stata Journal.

adopath
ado uninstall st0193
ado uninstall dea

will uninstall both versions of dea that are or have previously been available for network installation (as of the date this post went up).

[William Lisowski]

Let's improve further on the uninstallation code in post #6.

adopath
ado uninstall st0193
ado uninstall dea
which dea
which malmq

where the which commands should confirm that there is no other copy of dea.ado or malmq.ado to be found.

[William Lisowski]

The Sourceforge software release of the dea command referenced in posts #4 and #5 has an error discussed in the topic at

https://www.statalist.org/forums/for...script-invalid

Under some circumstances using the saving() option on the dea command causes a "syntax error" message in reaction to some internal command within dea.ado (not to the syntax of the dea command the user issued).

The workaround is that rather than using the saving() option, the output of help dea tells us that dea returns the results matrix in r(dearslt). This will be a standard Stata matrix that you can access using the Stata's matrix (not Mata!) commands documented in the output of help matrix and all the help files linked from that output. For example,

matriix result = r(dearslt)
matrix list result, noblank nohalf noheader f(%9.6g)

will copy the results matrix from its temporary location and then reproduce the tabular output dea placed in its log and offered to save.

[William Lisowski]

At the time of this posting, there is a malmq2 package both at SSC and at the developer's github site at https://kerrydu.github.io.

Both copies are identifed as Version 2.1 and show the release date of 20 November 2019 in the ado file.

However, the version at SSC has had a problem described at this Statalist topic which reports that the version at github fixes the problem.

At this time, the version at github can be installed with

net install malmq2, from("https://raw.githubusercontent.com/kerrydu/malmq2/master/")

but it is probably best to follow the link to the github site and confirm this before attempting to install it. Clicking on the malmq2 link there, and scrolling to the bottom gives the installation command.

I note that from inspection of the malmq2.ado file it appears that the malmq2 command does not require the dea command to be installed.

[Saloni Khurana]

Thanks, William, for creating this thread.
I am using malmq2 command, it is providing me with changes in efficiency. Is there any way to extract efficiency scores?

[William Lisowski]

I regret to admit that my knowledge of the Malmquist index is limited to having repeatedly figured out where to find and how to install the latest version of this software, I do not use it myself, nor do I understand the methodology in implements.

[Juan Price Elton]

Hello everyone: I am working with the command malmq (Malmquist Index) and wonder how I can use the option "variable Returns to Scale" ... I tried

xtset i t
gen dmu=i
tostring dmu, replace
do ldea.do
set matsize 11000
malmq input_1 impute_2 = output, ort(i) period(t) sav(file_input_oriented) replace

and it worked well
as far as I understand the option CRS (constant returns to scale) is the option by default. If I try to add the option rts(vrs) like as follows (I omitt the first lines so as to avoid an unnecessary long message)

malmq input_1 impute_2 = output, ort(i) period(t) rts(vrs) sav(file_input_oriented) replace

it doesn't work

I'd appreciate if any of you lend me a hand.
Thank you very much

[Juan Price Elton]

Hi: I already solved my previous doubt. What I was asking is basically non-sense. The Malmquist index doesn't make any assumption regarding the returns to scale of the underlying technology. Cheers

[Juan Price Elton]

Hi everyone:
I am estimating the Malmquist Index with the command malmq. My question is simple: when before using the command malmq I run ldea.do (something that must be done) does this do file estimate the DEA scores based on a technology with constant returns to scale, variable returns to scale or both?
Many thanks, regards and happy 2021
Juan

[Aline Aquino]

Hello everyone,

Data Envelopment Analysis allows you to use weight restriction to discriminate whether a particular output or input is more important. Given this, I would like to know if it is possible to use weight restriction when calculating the Malmquist Index in Stata?
I know there is another program that is possible to add weight restriction, but it doesn't decompose efficiency, and I need that decomposition.

If anyone knows, I will appreciate the help.
Thanks,

Aline