gravity.distances

The R package gravity.distances provides aggregate distances between geographic entities — such as countries, US states or Canadian provinces — that are consistent with the gravity equation in international economics. Please check out julianhinz.com/data/#gravity.distances for more information.

Installation

Install from Github via the devtools package:

devtools::install_github("julianhinz/gravity.distances")

Examples

The get_distance function provides distances between two geographic entities for a given point in time for a given value θ. The θ is a parameter in the aggregation of the mean that yields the harmonic mean for θ = -1, the geometric mean for θ = 0 and the arithmetic mean for θ = 1. In almost all cases in which a general gravity relationship is assumed to hold, i.e. when the underlying data generating process posits a negative relationship of the variable of interest with distance, the harmonic mean should be used.

Distance between two geographic entities

To get the distance between two countries, simply set the origin and destination argument to the ISO 3166-1 alpha-3 country codes of the two countries. The year argument defaults to 2012, the theta argument defaults to -1, i.e. the harmonic mean.

library(gravity.distances)

get_distance("DEU", "CAN")
# [1] 6519.294

Distance over time

Specifying an origin, destination and year delivers the harmonic mean distance (θ = -1) between to countries, here between Germany and Canada for the years between 1992 and 2012.

library(ggplot2)

dist <- data.frame(origin = "DEU", destination = "CAN", year = c(1992:2012))
dist$distance <- get_distance(origin = dist$origin,
                             destination = dist$destination,
                             year = dist$year)

ggplot(dist) +
  theme_minimal() +
  geom_line(aes(x = year, y = distance))

Distance by theta

Specifying the thetas for a given year shows the effect of θ on the aggregate distance. The result is most visible for short distances, e.g. the average distance between two points in Canada. Note that the data argument needs to be set to distances_from_countries_to_countries in order to have distances for theta values between -2 and 1 in 0.1 increments.

dist <- data.frame(origin = "CAN", destination = "CAN", theta = c(-20:10)/10)
dist$distance <- get_distance(origin = dist$origin,
                             destination = dist$destination,
                             theta = dist$theta,
                             data = "distances_from_countries_to_countries")

Distance between other geographic entities

The data argument can be used to use different distance datasets.

dist <- expand.grid(origin = "BC", destination = c("BC", "AB", "SK", "MB"), theta = c(-20:10)/10)
dist$distance <- get_distance(origin = dist$origin,
                             destination = dist$destination,
                             theta = dist$theta,
                             data = "distances_from_canada_provinces_to_canada_provinces",
                             data_store = F)

Further options for get_distance

• data_store can be set to FALSE in order not to store additional downloaded datasets from the gravity.distances_data repository. Default is TRUE.
• data_url may be specified for custom datasets. Default is https://github.com/julianhinz/gravity.distances_data.
• code_format may be set to specify the format of origin and destination in accordance with the countrycode package. Default is iso3c.

Other functions

The remove_data function removes some or all downloaded additional datasets. The data argument can be set to a specific dataset to be removed, e.g. data = distances_from_canada_provinces_to_canada_provinces. The default is data = NULL, deleting all previously downloaded additional datasets.

Roadmap

• determine required dataset automatically
• fix geometric distances for US states and Canadian provinces
• potentially extend to Python and or Julia, foundation laid with feather file-format for additional datasets

Bugs?

• This package is still in its very early stages. Let me know if you find bugs via pull request or e-mail to mail@julianhinz.com.