http://www.lboro.ac.uk/gawc/visual/hwatlas.html
1. We have selected the top 123 cities in terms of global network connectivity and measure the connectivity between each city and the other 122 cities.

2. These individual city inter-linkages closely reflect the overall pattern of global network connectivity (i.e. every city is most connected to either London or New York).

3. For each city we regress its connectivities with other cities against their global network connectivity.

4. We compute the residuals from the regression line and these are interpreted as a city being either over-linked (positive residuals) to another city (relative to its global network connectivity) or under-linked (negative residuals). It is these residuals that are mapped to show a city’s hinterworld.

5. The standard error of estimate indicates how close the scatter of points is to the line: it is interpreted in this context of indicating the specificity of a city’s hinterworld (i.e. zero would indicate a hinterworld exactly the same as the global network connectivity pattern and therefore the higher the standard error the more unlike a city’s hinterworld is from the general pattern of global connectivities. In the analyses, standard errors range from 0.013 (Madrid) to 0.067 (Indianapolis)).

6. The same ordinal scale is used for all maps to facilitate simple comparisons between hinterwords. An initial discussion of comparisons can be found in chapter 5 of P.J. Taylor (2004) World City Network: a Global Urban Analysis (London: Routledge).

7. The maps are in cartogram form to show each city in equally. The cartogram places cities in their approximate relative geographical positions.



Most over-linked city: Prague (0.102)
Most under-linked city: Chicago (-0.106)
Specificity of hinterworld: 0.043