Measuring Digital Inequality

High-resolution population estimates across 147 countries and base station distributions all over the world reveal worldwide digital divide.

We propose a fine-grained telecommunication service inequality index for any given geographical region, by using only the population and number of base station information.

Take the Republic of Tunisia as an example, the calculated region-wise inequality indexes are displayed in the below figure. From this figure, one can clearly and vividly observe the inequality index of each region across the whole country. Some regions such as those near the capital Tunis have very low inequality, which indicates good connectivity. Efforts of mobile network operators enrapt in improving the connectivity of these regions are unavailing. On the contrary, they should be attentive to regions with high inequality as few resource deployment are likely to bring huge social and economic benefits.

Inequality Index visualization for Tunisia (Grid-wise).

Inequality Index visualization for USA (Grid-wise).

Inequality Index visualization for France (Grid-wise).

Inequality Index visualization for Saudi Arabia (Grid-wise).

Inequality Index visualization for Uganda (Grid-wise).

We validate the effectiveness of our inequality index by comparing it with a coarse-grained measurement, i.e., the GSMA mobile connectivity index. The derived Pearson correlation coefficient is 0.92, indicating a high correlation exists between our index and the GSMA index. The achieved statistical p value is 2.2e-16, which is small enough to reject the null hypothesis (without correlation) and embrace the alternative hypothesis (with true correlation).

We can conclude that our index is consistent with coarse-grained measures, indicating the effectiveness of our proposed telecommunication service inequality. However, compared with the GSMA index that provides only country-wise statistics, our index is much more fine-grained and can provide inequality statistics for any given geographical segment.