Pankaj Ghemawat and Steven Altman (PDF):

While of obvious appeal to certain urban elites, this picture turns out to be factually wrong. While New York is usually rated as one of the world’s top global cities, prior research using Sassen’s preferred measure indicates that New York’s greatest connectivity is with Washington, DC, ahead of Tokyo, and Chicago and Boston round out its top four con- nections. Other US cities are much less connected inter- nationally: thus, the Los Angeles metro area, the fourth largest in the world in GDP terms after New York, Tokyo, and London,12 counts only one foreign city (Tokyo, at #8) among its top dozen connections.13

Thus, even as long-term trends point to the rising impor- tance of global cities, there is evidence that cities—like countries—conform to the laws of globalization that were articulated in the conclusion of Chapter 1.14 Paralleling the law of semiglobalization, flows often take place more intensively within large cities than between them. For an example pertaining to trade, the value of shipments within a given zip (postal) code in the US (with a median radius of just four miles) is three times larger than the value of shipments across zip code boundaries.15 And in regard to capital flows, investment fund managers are more likely to buy or sell stocks when other managers in the same city are doing so.16

Both laws of globalization are also in evidence when one looks at patterns of who follows whom on Twitter. Over- all, 39% of all Twitter ties turn out to be local as in within the same (roughly metropolitan) regional cluster, 36% fall outside the regional cluster but within the same country, and 25% are international (as we noted in Chapter 1). Nor do these average tendencies necessarily weaken with city size. Thus, in Sao Paulo, one of the biggest hubs of Twitter activity in the world, more than 75% of the ties were local!17 And Figure 3.4 highlights how Twitter ties drop off with physical distance. This analysis of Twitter also backstops the earlier point that even supposedly global cities still tend to be more connected to their domestic hinterlands than to other cities abroad. Figure 3.4 indicates that the overall pattern of extreme distance-dependence is affected notice- ably only by a spike at the New York–Los Angeles distance (a domestic link); New York–London is just a blip, if that, in the overall pattern, and the other city pairs highlighted in the figure have no discernible effect at all.