“Grave-robbers cut away part of a false door bearing painted stone reliefs depicting
Data on incidents of armed conflict were compiled from two sources of event data: the Armed
the ACLED record information on a variety of political violence incidents in Egypt from 1997 to 2014,
including three types of battles, violence against citizens, rioting, protesting, and non-violent events
around the world from 1970 to 2014. To be included, an incident must be “an intentional act of violence
included if they meet at least two of the following three criteria: (1) the violent act was aimed at
attaining a political, economic, religious, or social goal; (2) the violent act included evidence of an
intention to coerce, intimidate, or convey some other message to a larger audience(s) other than the
Arts
2018, 7, 22
9 of 26
immediate victims; and (3) the violent act was outside the precepts of international Humanitarian
Law (
LaFree et al. 2015
, pp. 19–20). The data from both databases were filtered to include variables
on the incident date, country, and location of the incident (at the governorate, city, and site level).
The data were cleaned to remove any duplicate events both within and between data sources. In total,
there were 5762 incidents of armed conflict in Egypt from 1997 to 2014. During the time period in
question, there were several regime changes in the Egyptian government. Such changes can influence
both armed conflict and antiquities looting. As such, the study includes a binary variable indicating
the occurrence of a regime change. These data came from the Polity IV project, a part of the Integrated
Nations Center for Societal Conflict Research’s (INCSR’s) database. The polity data continually track
and update on regime changes around the world (
Center for Systemic Peace 2015
). Once all data were
cleaned and coded, they were aggregated and merged into a month-level dataset and a quarter-level
dataset (see Tables S1 and S2 in the supplementary files). These were the most granular units of
analysis for which there was sufficient variation in the antiquities looting and armed conflict variables.
Benefits and Limitations of the Data
These data have several benefits and limitations that are important to discuss. All sources of
data used in the current study come from open sources. Open source data have several benefits for
studying phenomena that have historically been easier to access through more qualitative methods.
First, by virtue of being publicly available, open source data are useful for studying new areas within a
discipline like criminology or archaeology. Generally, less data is available on such phenomena because
they have not been studied. For phenomena that are considered outside the traditional scope of research,
open source data provide a way to look at these new areas. Second, open source data are a cost-effective
way to collect data on a wide variety of subjects (e.g., terrorism, see
Dugan and Chenoweth 2013
).
Online digitization, publishing, and archiving of newspapers, journals, and blogs provides easy access
to decades of news articles from media outlets around the globe. Repositories can be specific to
a single institution (e.g., Reuter’s archives) or be large databases covering many large and small
publications (e.g., Lexis Nexis). Access to these databases through universities or private subscriptions
allows researchers to access large quantities of information spanning any subject. For example,
the GTD is considered one of the most robust terrorism databases currently in use in criminology
(
LaFree and Dugan 2007
).
Despite the utility of open source data, the sources used to create these data present some
limitations. First, while these data include a relatively large number of regime changes in the dataset,
they are still relatively rare in the data, which may affect the findings.
Second and more important to the analysis, the news stories used to create the antiquities
looting data contain unavoidable implicit bias. The current study could only capture what the
media chose to cover on antiquities looting, which changes over time. This includes what the media
considers to be newsworthy, what is of interest to the public, and the means of reporting information.
For example, the advent of the Internet made it significantly easier for journalists and amateur reporters
to disseminate information. This in turn broadened the range of newsworthy topics, making it more
likely that antiquities looting would be reported later in the timeline. Additionally, news stories may
lack granularity to get at the actual behavior of interest—antiquities looting. More dramatic or serious
cases of looting are more likely to be reported by news agencies, while every day looting may go
unnoticed or unreported. As such, the events in the data may disproportionately represent targeted or
strategic lootings compared to opportunistic lootings. The current study must also assume that objects
have been removed from the sites being reported as “looted”, which may or may not be a reasonable
assumption. As such, with these data, the closest this study gets is reports of antiquities looting.
2.2. Methodology
In addition to looking at antiquities looting and armed conflict descriptively, the current study
used multiple time series analysis to examine three hypotheses: (1) antiquities looting and armed
Arts
2018, 7, 22
10 of 26
conflict have a positive statistically significant relationship; (2) an increase in antiquities looting will
precede an increase in armed conflict; and (3) an increase in armed conflict will precede an increase in
antiquities looting. Specifically, this study uses autoregressive distributed lag models (ARDL) and a
bounds testing approach to look at each hypothesis.
ARDL models are designed to look at autoregressive processes, phenomena that are explained in part
by their own history and in part by the influence of other factors. In this case, the model allows us to look
at the influence of prior incidents of armed conflicts and of looting on current incidents of armed conflict.
The basic ARDL model is presented in Equation (1), where
∑ β
i
∆y
t−i
+
∑ β
j
∆x
1t−j
+
∑ β
k
∆x
2t−k
estimate each set of parameters in levels and
θ
0
y
t−1
+
Dostları ilə paylaş: