Mobility Files

Sources of data

Going out

econStatusMovementProb_week[day/end]_[single/multi].txt

Extracted based on Round 9 Manicaland survey as a function of whether or not someone goes out and has contacts within the community space on any given day. Used to reflect the likelihood that a person comes into contact with people outside their home.

Movement between districts

Mobility is being drawn from Call Detail Records (CDRs), which capture when someone has moved between locations based on which cell towers they access. Thus, a data point is only generated when someone makes or receives a call. The data is anonymised, meaning that we do not (and cannot) track an individual moving beyond the individual movements. That is, if someone travels from A to C through B and makes calls in all three, the recorded movements will be A -> B and B -> C, rather than A -> C.

Further, we cannot tell whether the same person has moved between locations multiple times, or whether multiple people have passed through a single time. We want to understand how people move between spaces, so this poses a significant challenge. We adopt a number of preprocessing steps to address this.

Preprocessing the CDR data

The CDR data is preprocessed in Stata; its intermediate form is stored in data/preprocessed/mobility/New Files.

See more at this GitHub for information about the different indicators.

Feeding the data into the simulation

There are a number of ways data is fed into the simulation

Making contact with people outside the household

We assume that individuals are more or less likely to travel beyond their immediate home based on their daily activities, which we in turn assume are based off of their economic status. Thus, someone who is a student is expected to leave the house every week day for school; someone who has disabilities or is out of work is expected to leave the home less frequently. This data is contained within the file econStatusMovementProb_week[day/end]_[single/multi].txt

Choosing a destination

The simulation expects 2 input files which control the destinations of those travelling outside of their home locations:

1. Prelockdown District movement file: daily_region_transition_probability_new-district_pre-lockdown_i5.csv
2. Lockdown District movement file: daily_region_transition_probability_new-district_post-lockdown_i5.csv

These files follow the same format of a mobility matrix which provides the probability of individuals from each of the districts (d_1 - d_60) staying in their home district or traveling to any other district. This probability is provided for each day of the week (days are numbered 0-6, 0 being Monday). File 1 shows higher rates of movement than File 2 due to the movement reductions among districts during the lockdown period.

These files can be used to impose travel restrictions within the model. For example, if all travel to or from a certain district is banned, the probability of an individual travelling to or from this district would be 0.

Travel during lockdowns

The simulation can, theoretically, incorporate the decreased likelihood of travel during the lockdown. That is, we could include not only the change in destinations (through the district movement files) but also the decreased chances of leaving one's home district. Thus, economic activity might still prompt the person to leave their home, but they are less likely to leave their community during a lockdown. This data is stored in the Intra-district mobility rates reduction file intra_district_decreased_mobility_rates.csv

Putting it all together

Together, then, the individual makes a choice as follows:

1. Leave home (econStatusMoveProb)
2. If leaving home, choose the target destination: which district