In 2019 Kieran Baker and Marco Dosi were awarded a Small Grant for Research Students by the Faculty of Arts and Humanities, King’s College London, to perform statistical tests (e.g. algorithmic processing, logistic regression analysis) on the consular dataset, which culminated in the development of a geospatial model that is able to predict the expected dissemination time of consular announcements.
The DataCons Geospatial Model is now under final testing and the team is preparing an interactive map which will enable users to explore its results.

More about the Model

The model innovatively applies the Cox Proportional Hazards model, a time-to-event statistical approach traditionally used in the modelling of survival times and understanding of risk factors in medical contexts - a problem known as Survival Analysis. This model is uniquely adapted to handle spatio-temporal data, particularly in scenarios where the exact timing of an event is uncertain within a window of time. There are three distinct models, each varying in the scale of geographic specificity—from the precise location of the consul’s announcement to broader regional and macro-regional contexts. These models are adept at handling varying levels of data availability, with the first model being most effective in data-rich environments and the latter two providing robust estimates in data-deficient scenarios.

A key innovation of this approach lies in its ability to work with censored data, a term in statistical analysis referring to incomplete data where some entries are only partially observed or are missing crucial elements. This is a common challenge in historical datasets where records may not be fully preserved. The model categorises consular data into three types: post-consular, consular with partial information, and consular with complete information, each representing different stages of information dissemination and types of censoring. The models incorporate a range of explanatory variables, including the year of the event, starting location, estimated travel times from ORBIS, and the nature of the journey (single or two-legged), among others. These variables are used to adjust the survival curve for each journey, providing nuanced estimations of dissemination times.

The application of geospatial modelling in this context is groundbreaking. It not only offers a method to predict dissemination times based on historical data but also opens avenues for understanding the dynamics of information flow, movement, geography, space, and the working of the bureaucracy in ancient settings. This model’s potential extends beyond historical analysis, offering insights into modern scenarios where data is incomplete or censored, and the spatial component is critical. The study sets a precedent for the use of advanced statistical models in historical and geographical research, demonstrating the untapped potential of integrating time-to-event analysis with geospatial data.

Graphs below show an example of Travel-Time Range vs. Dissemination-Time Range. Data is representative for the trajectory Italy to Egypt in the years CE 311-541 (Dosi 2022: 479—80)

	311-380	381-541
minimum	1.950092	2.427591
median	2.077409	2.652893
maximum	2.091131	3.050766
mean	2.039544	2.681835

The mean ratio of dissemination time to travel time is 2.039544 [1.951538 - 2.127550 95% Confidence interval for the mean] for the period 311-380, and 2.681835 [2.626471 - 2.737198 95% CI] for 381-541.

Data indicates that the time taken for consular announcements to disseminate was, on average, about twice as long as the actual travel time. This significant delay suggests that dissemination was not a direct process. Instead, it likely involved intermediary actions or challenges, such as logistical constraints, administrative procedures, or even intentional delays in communication. Understanding these delays helps highlight the complexity and inefficiency of information flow in the late Roman period, offering insights into the bureaucratic and social structures of the time.