Minutes of the MID3 Meeting, held at Penn State University, July 7, 2001

Stuart Bremer chaired the meeting. In attendance were Archana Bhandari (from MID-Central), Paul Diehl (Far East), Daniel Geller (South Asia), Mark Gibler (Africa), Charles Gochman (Middle East), Paul Hensel (South America), Glenn Palmer (Central and Eastern Europe), Brian Pollins (North America), James Ray (Africa), Patrick Regan (Oceania), and Richard Stoll (Western Europe)

Interim Report on Status of Project

1. Bremer expressed happiness with the Forum on the website and (http://mid3.la.psu.edu/) said it may ultimately be included as part of the data documentation.
2. He reported that all regions are in pretty good shape except for the Middle East. Gochman reports all incidents have been coded through 2000 and will be processed and entered hopefully by the end of August. Regarding Tel Aviv’s contribution: Penn State will take 1998-2000 for those. By the end of August, 2001, all incidents through the end of 2000 will be entered.
3. Bremer reported that 1268 incidents have been entered thus far; 980 have been OKed by MID Central; 97 have been deleted; 92 to be reviewed. He estimates that 15-20% are duplicates, which will be discovered in the MID generation phase.
4. Re dispute generation, 977 incidents have been assigned to proto-MIDs. Proto-MIDs have been assigned and some entered, but none been reviewed just yet. MID Central will slow down a bit this summer. Bremer believes 80% of the MIDs will be easy, but expects problems in the Middle East, Central Europe and Africa.
5. Budgets: If there are resources at the regional centers left over, we may ask NSF to consolidate the funds from the various institutions to PSU for final preparation of data. Miscellaneous stuff will doubtless remain, e.g., documentation. If money is needed at the regional places, spend it, but if there’s money left over, don’t splurge. PSU may need it. Paul Diehl will talk to Frank Scioli about this.
6. Revised work schedule:
   Construction of narratives: Stuart wants stories for each incident as well as MIDs. There is variance across regions about how detailed the story is. Bremer will create a new field on incident entry form: STORY.
   MID narratives. This should not be simply a summary of the incidents, as the MID is bigger, and requires a different perspective. We would like to see references in the story to various specific incidents. This raises the issue of the numbering system: how do you refer to other incidents? It was decided that the current incident numbering system may become permanent. Permanent MID numbers will come later. The information placed in the new STORY section may be taken from the current NOTES section. NOTES should be for ancillary stuff, e.g., why the end date is what it is, if the beginning date is unclear, etc. Duplicate information in the story box and the notes is OK.

Incident Collection Issues

1. We discussed the issue of whether there be a final check/search for incidents we may have missed. There was concern that we may be too Lexus-Nexus oriented. It was noted that FBIS is online. It may not be available to everyone. Cross-regional checking during the MID aggregation process may be a good idea. Bremer argued that looking for new incidents may not be useful, and widespread checks may not be a good idea. We don’t want to spend a lot of time going back over old territory. Ray said he was now less enthusiastic about doing this than he had been, as it’s not a rational allocation of resources. Bremer concluded that while new incidents will be found in the course of constructing MIDs, we will not undertake a serious new search.
2. Routine incidents: This focused on events that happen so frequently it has become routine. Stoll observed that routine interactions (Iraq no-fly zone) are difficult to code because they are not reported. Bremer said we should code what we find, and suggested that at times a note may be put in that says, in effect, “This type of thing has been going on a long time, but we lack precise information to allow coding.” But in the absence of information, we can do no more than say, for example, “This was the 14th time this has happened.” But we can’t code what we don’t know.
3. Target-actor: Question was raised about KFOR in Yugoslavia. Palmer and Stoll said that a YUG threat against KFOR is directed against all states that are present, not those that are members of the organization. This was accepted. Ray asked about a situation where A sends troops to B, arguing they’re part of a peacekeeping operation, but B rejects that. Is that an incident? It was decided it was.
4. Credibility of reports and reliability of incidents need to be higher if it becomes a one-incident MID. Judge the context and plausibility of information.
5. Start-end dates: Hensel asked about ending dates for seizures. Bremer said that MID Central has been moving towards a 3-day rule if there’s no information about the release. Gochman wondered about border violations when we lack information about when the violators leave. Similarly for border fortifications, mobilizations, shows of force, and occupations. Whatever information is available should be in the notes. Hensel noted that for MID2, seizures and occupations are coded as lasting only 6-months if they lasted at least 6-months. Stoll argued that we need to adopt some rules about this. MID Central will come up with the rules, and do so by observing difficult cases.
6. Incident-type distinctions: Geller made an argument that while we do not usually code weapons tests as incidents, Pakistani nuclear test should count as a show of force. After some discussion, MID Central asked Geller to write a long note to justify the inclusion of this particular test is an incident.
7. Incident-type distinctions (II): Gochman asked about TUR incursions into IRQ to go after PKK rebels. ME originally coded this as an occupation, not an attack because the target isn’t the state, but rebels on territory of another state. MID Central asserted that this should be an attack. That position was accepted.
8. Incident-type distinctions (III): There are frequent attacks on Hezbollah forces by Israel, and perhaps even on “Lebanese citizens”; those types of incidents only count if there’s a Lebanese protest.
9. Issues: Hensel said that to be consistent with past efforts, maritime issues should “policy.” Fishing rights or oil rights are policy issues. Territorial disputes are dry land disputes, but if the issue is one of water, it’s “policy.” You may include multiple issues. This is particularly critical for MIDs, but we are enjoined from going back and to recode incidents. Hensel will provide a clarification of coding rules regarding issue type for MIDs.
10. Issues (II): Complicated situations involving Kenya and Sudan surround rebel forces in each other’s territory. The issue at stake, we decided, is defined by who’s trying to alter the status quo. Forces in Ethiopia and Eritrea are looking for land. There’s no clear rule about how to define the issue. The question needs to be researched.
11. Fatalities: Missing v. zero: If there is a reasonable possibility of deaths in an incident, but you have no information, then "missing" is the correct code. If the nature of the act makes it impossible or very unlikely that deaths occurred, then it’s zero. Or if you know that no deaths occurred, e.g., threats, it’s zero.
12. Finalizing the incident data: MID Central wants to seal incident data base sooner rather than later.
13. Pollins asked whether we should we edit incidents as we go through the MID aggregation process and find new information. Yes. Regional center should email MID Central that they have changed previously OKed incidents. Additionally, we’re now swapping incidents, and if you make changes in OKed incidents you should cc MID Central and the region that now has the incident. The computer program has the date of the last edited.
14. Lexus-Nexus has a lag time before it records new stories. How long should we wait before we use it as definitive, particularly for 2001 incidents? Bremer thinks a lag of about 3 months is sufficient. Stoll noted that Western Europe has not faced any problems in its 2001 incident search. Hensel says Latin America has experienced no problems, and estimates the lag to be 2-3 days.

Dispute Generation Issues

1. Demarcation Problems: If the regions have a proto-MID that is in fact 2 or more, then alert MID Central and they will assign new MID numbers. Let them know how the proto-MID is being split. The MIDs provided by MID Central are just a first-cut. MIDs can be distinguished by temporal proximity, location, behavior, etc. Issue coding should probably not be used, as it’s not as reliable as other information, and issues can change during a dispute. If they begin as two separate MIDs they cannot be merged, but should be “related.” But one MID may break into two or more, but you should make clear in the Notes why this was done. Can a dyad have more than one MID simultaneously? Yes, as the US-USSR dyad demonstrated. You should cross-reference MIDs, that is, note that MIDs are related or tied in some fashion, e.g., the two no-fly zones in IRQ, does that lead to two MIDs or one? Probably two that are related.
2. MID termination after war. Bremer pointed out that in the previous MID effort (where only escalatory incidendents were coded) war was the highest level of hostility possible; hence a MID that went to war ended when the war ended. However, in the MID3 effort it is possible for a qualified incident to occur (and merit coding) shortly after the end of a war. It is unclear in these instances whether and when these post-war incidents should be considered part of the continuing MID-war or the start of a new MID. If they are considered part of the MID-war, then the MID end date would be reckoned from the date of the last incident, not the war end date.
3. End dates: MIDs end when the last incident ends. No incidents for 6 months, you then backdate the MID to the last date of the last incident, as per page 2 of MID coding rules.
4. Participants, state and non-state actors: MID narratives should include reference to the relevant non-state actors, both sub-national and international. International organizations, rebel groups, organized resistance, terrorist groups, ethnic minorities, etc. They must be involved “in some important way”, i.e., they have exerted some military influence. In other words, if they had been states, their involvement would have been codable. Don’t include the non-state actors if they’re the targets, only if they’re the actors. We should let the person in charge of the proto-MID know that there were non-state actors involved. You need to know who has your incidents, and when you know that, send them the information.
5. Related incidents: they refer to those in this particular MID. The current incident numbers are the permanent incident numbers.
6. Pollins asked how incidents are included in the proto-MIDs? Bhandari does it, and provided a brief description of how she decides. If there are questions, email Archana with queries.
7. Duplicate and similar incidents: Duplicate: they’re identical and one of them should be deleted. MID Central should be informed that one of them should be deleted. Two that are slightly different: discussions should be opened by the center with the proto-MID and the initial coding region. The MID people are the ones who are in charge, and they should inform MID Central of any decisions. If you think something is in fact not an incident, you should similarly discuss it with the initial center, and then ask MID Central to delete the incident, if that’s the decision. Ray asked how MIDs are allocated. MID Central pays attention to a variety of factors, including the creation of an equitable workload.
8. Outcome and settlement: Palmer supplied new wording of the outcome. Outcome applies to the “side” coding, not to the participant coding. Diehl asked how far into the future can you look do determine the outcome and settlement. Hensel said that MID2 looked at the sequence by which the MID ended (or what Bremer calls the “immediate outcome”) rather than the long-term or ultimate outcome. We should ask Phil Schaefer about how that was done (what time frame was used to identify a dispute’s outcome/settlement). 6-month rule implies to GP that 6 months is when you should make the decision.
   [Phil Schafer confirmed  Hensel's recollection that the it is the "immediate outcome" that is assessed. Hence, for the sake of consistency with MID2, this rule will be followed. There may be an additional outcome variable added to satisfy those who are unhappy with this rule. See the MID3 forum for further information. SAB]
9. Fatalities: Sometimes we will have to make estimates. To do this, take into consideration the nature of the action, the 3:1 casualty:fatality generalization, etc. Reasoned judgment needs to be exercised.
10. Participation dates: End date of the last incident for which the specific state was involved. You just take them right from the incident sheets, though other research may be required.
11. Revisionist states and issues. Pollins worries about the difference between the way the literature defines “revisionist” and the way we do it. Our meaning is situation, and does not refer to the global system. The status quo is that which exists on the date the dispute begins. Should note what the revisionist state is seeking. MID Central: Everyone should include a description of the issue in the issue note. Use Note #9.
12. Links: This is going to be difficult, because we won’t know all the final MID numbers until all the work is done. It will probably have to be done iteratively. Links to wars will be particularly hard, as the war list may be evolving throughout our research.  In general, links should be direct, not long-term, and follow in a direct way from pervious dispute. Hence, they should be temporarily proximate.
13. MID Central requests extensive use of the Notes to provide information on decisions and narrative.
14. The finalization of the MIDs will be a difficult process. There may be lots of checking and rechecking as things are cleaned up.

Other Matters