Pre-1993 MID Data

Dyadic MID data is only available from the Correlates of War project from 1993-2001, because incident data (showing direct dyadic interaction) has only been systematically coded since 1993.  For the pre-1993 period, either assumptions must be made in coding dyadic interaction (critically, that all states on opposite sides interacted), or additional research must be performed to identify actual interactions.  In addition, coding to a dyadic form must carefully check dates of state involvement to verify that states were opponents at the same time.

MID v.2.1 (running from 1816-1992):

MID v.2.1EE provides information on a subset of escalatory incidents in 1816-1992:

There are two other versions of the dyadic MID data available for the pre-1992 period which tackle these issues in different ways. Extensive documentation on procedures are available in the documentation at both sites:

Scott Bennett’s EUGene software (v3.2, 2008):

Zeev Maoz’s dyadic MID data (v2.0, 2005):

Updates

1. Data Updates for MID v.3.10:

   On September 26, 2007, the MID data were updated to version 3.10. Updating for version 3.10 focused on the following issues:

   • Since the release of v3.02, the COW project has investigated error reports, questions, and reports of discrepancies on nearly 150 particular MIDs submitted by numerous scholars, including Zeev Maoz, Patrick Regan, Russell Leng, Ashley Leeds, Mark Souva, and others. Each report has been investigated, and v3.10 has made changes/corrections to 74 different MIDs (the remaining MIDs did not require changes; error reports typically resulted from a misinterpretation of coding rules). There have been changes to a total of 100 MIDs since the release of MID 3.0. You may review a detailed cumulative list of these questions and changes that resulted here. A detailed listing of questions examined where it was determined that no changes in the data were required is here.
   • Dyadic MIDs: The COW dyadic MID set (1993-2001) breaks down overall MIDs, some of which have many participants, into actual disputing pairs of states. Each dyadic MID has exactly two states involved on opposite sides; each was directly involved in militarized incidents against the other. The data set omits dyads where the two states were on opposite sides of a multi-party dispute, but never took action towards one another. While there are approximately 300 MIDs from 1993-2001, there are over 500 dyadic MIDs during this period. Version 3.10 corrects errors in the dyadic data set. These errors most often resulted from errors in the underlying (non-dyadic) MID data sets that carried over to the dyadic MID set. Other corrections and additions to the data include the following. 1) Dyadic MIDs and the related dates were not being correctly computed for states exiting and reentering a MID; in such circumstances, a new dyadic MID is now correctly coded. 2) The data now include a "dyadic MID number" as well as the base MID number; this number separates each pair of disputing states, and separates incidents of repeated involvement. 3) Hostility levels were not always reported correctly in the dyadic MID data set in the case of war. It is sometimes the case that a MID has a hostility level of "5" (war) even though no single incident reaches hostility level 5. This is because multiple instances of a use of force (hostility level "4") can accumulate to a level of war. These cases were not always picked up in the MID 3.02 data; they have been corrected with hostility levels in such cases being taken from the MIDB data set rather than the MID IP (incident) data. 4) The end dates of dyadic MID were in some cases not being correctly computed; the end date of the last incident between two states in a MID is not automatically the correct end date of the dyadic MID, as MIDs can end days or weeks after the end of the last incident. These dates have been corrected.

2. Incident Level Data and Dyadic Data Updates for MID v.3.02

   On October 10, 2003, we replaced version 3.01 of the MID data with version 3.02. The data sets for version 3.02 include the following:

   • Incident and Incident Participant level data files: Incidents are the building blocks of MIDs, and represent individual militarized actions incidents (specific threats, displays, or uses of force) that comprise the disputes. MIDs may consist of one or many specific incidents. The MID 3.0 incident data collection reports information on each of the militarized incidents occurring with MIDs (incidents may be either escalatory or de-escalatory). MID-I contains the essential elements of each militarized interstate incident from 1993-/2001 (incidents that belonged to disputes that began in 1992 and continued into 1993 were also collected). MID-IP describes the participants in each of those incidents.
   • Dyadic MIDs: The COW dyadic MID set from 1993-2001 breaks down overall MIDs, some of which have many participants, into actual disputing pairs of states. Each dyadic MID has exactly two states involved on opposite sides; each was directly involved in militarized incidents against the other. The data set omits dyads where the two states were on opposite sides of a multi-party dispute, but never took action towards one another. While there are approximately 300 MIDs from 1993-2001, there are over 500 dyadic MIDs during this period.
   • Minor Corrections: We have made a few minor changes to data on the MIDs that were newly collected between 1992 and 2003. You may review a summary of these changes here.
   • Revisiting older MIDs: In response to various comments and questions about MID codings from the v2.1 data set, we have revisited a number of pre-1992 MIDs and made 25 corrections/changes/additions. You may review a detailed list of these questions and resulting changes here, and a detailed listing of questions considered which did not necessitate changes here.

   Because the new data version include numerous documentation files, we have released all documentation files in .pdf format only. Users wishing to cut and paste from these documents can still do so using the "t" keystroke command within Adobe.

   A typo was discovered in the MIDA_3.02.csv data file on October 14, 2003 (the MID number for MID 21 was mistakenly listed as a 20, so it appeared that MID 20 was listed twice and MID 21 not at all). The MID number for MID 21 has been corrected in MIDA_3.021.csv, which replaces MIDA_3.02.csv. No other data files were affected by the typo.

3. MIDB Dataset Updates (v.3.01)

   On March 23, 2003, we replaced version 3.00 of the MID data with version 3.01. The data in the MIDA file (main MID listing) is identical from version 3.00 to version 3.01. However, we found an erroneous case in version 3.00 of the MIDB file, and have deleted this case in version 3.01 (it should not have been in there in the first place). The case was MIDB data set dispute number 4241. This is the only change from version 3.00 to version 3.01.

   With this release, we have also made available text file versions of the documentation as .rtf files, along with .pdf files.

Development Materials

Version 3.0 was developed through a 3-year multi-university initiative supported by the National Science Foundation. The project would not have been completed without the participation of co-PIs Scott Bennett, Stuart Bremer, Paul Diehl, Dan Geller, Doug Gibler, Paul Hensel, Chuck Gochman, Zeev Maoz, Glenn Palmer, Brian Pollins, Jim Ray, Pat Regan, and Ric Stoll. We have preserved the key documents used during the project's data creation phase. To access these documents, or to review the participants in the project, select among the following links:

• Project Description
• Project Personnel
• Participating Institutions
• Project Documents
  

Questions and Feedback

The militarized dispute data are hosted by Glenn Palmer, Pennsylvania State University, under the COW Data Set Hosting Program. In case of questions or concerns concerning the data or coding rules, he may be contacted by email at gpalmer@psu.edu.