The history of long distance trade can be broadly divided into three periods: the globalisation or mass trade period (from the 1820s onwards), a transition period (in an integrating Atlantic economy, from the C16th onwards) and the pre-globalisation or network trade period (prior to the 1820s outside the Atlantic economy).
What drives the transition from one period to another is communication and transport costs. Prior to the development of steamships, railways and the telegraph, communication and transport costs were so high, one cannot talk with any seriousness of an economically globalised economy. There was simply no significant convergence of prices for commodities in different regions; so one cannot talk about global markets for, as economist Deirdre McCloskey points out:
The only relevant standard for “one market” is similarity of price.
Instead, there were series of trading networks linking local markets with divergent prices.
For example, as soon as significant trade connections were established between the Mediterranean world and China, the tendency was for silver to flow from the Mediterranean world to China, because there was no world price for silver; silver remained much more scarce compared to output in China than in the Mediterranean world, so silver flowed from where it was cheaper to where it was more expensive, and did that for roughly two millennia.
A PSFM world
Where specie (gold or silver) was the dominant medium of exchange, if you were buying more goods and services than you were selling (a negative trade balance), specie would flow out. If you were selling more goods and services than you were buying (positive trade balance), specie would flow in. In such a world, David Hume‘s price-specie flow mechanism (PSFM) makes sense: indeed, describes what happened.
So PSFM describes the pattern of Eurasian trade from at least Roman times onwards. The pattern was particularly intense under the late Roman Republic and Early Empire, when the Roman state had access to major silver mines. A key driver of the Crisis of the Third Century (235-284) was the exhaustion of said silver mines which massively undermined the trade flows that many states relied upon.
Said crisis was much wider than just Rome’s difficulties. In rapid succession the Han dynasty collapsed (220), the Parthian Empire was overthrown (224) by the House of Sasan, the Kushan empire declined (c.230), and states around the Indian Ocean (the main conduit for trade between the edges of Eurasia) declined (such as the supplanting of the early Tamil dynasties by the Kalabhras dynasty c.250). Indeed, the remarkable thing was that the Roman Empire survived, albeit profoundly institutionally altered. (There was, however, some continuity between the Parthian and Sasanian empires.)
The silver-for-goods pattern became intense again when the development of better pumps and silver-lead-copper smelting led to large increase in the output of Central European silver mines (pdf) from the late C15th to the early C16th followed by–with the looting of the Aztec and Incan empires and the discovery of the Potosi silver mountain in the C16th–a flood of American silver into what was now the Atlantic economy. The silver-output ratio in the Atlantic economy made Atlantic goods even more silver-dear, and Asian goods comparatively silver-cheap (and silver goods-cheap in the Atlantic ecomony and goods-expensive in the Asian economy).
Silver was the main medium of exchange in the Eurasian trade networks and Atlantic economy, as it had been for centuries in the Eurasian trade networks, so goods flowed from Asia (particularly China) into the Atlantic economy and silver flowed from it to Asia (particularly China). That was the main driver of trade in (what were now) global trade networks up until the 1820s.
The one major exception was sub-Saharan Africa, which mainly used gold as a medium of exchange and mainly exported slaves–to Islam (as it had done for centuries) and across the Atlantic, to the Americas significantly emptied of local labour by the disease catastrophe of the Columbian exchange. (I.e. the importing of the entire Eurasian disease complex effectively all at once to populations with no immunity.) So, to the horrors of the trans-Saharan slave trade was added the horrors of the trans-Atlantic slave trade.
The Atlantic transition
There were only relatively minor improvement in transport and communication technology up until steamships, railways and the telegraph (i.e, the 1820s). But, as military folk say, quantity has a quality all of its own. The Atlantic economy saw a massive increase in the scale of water-borne transport, both via canal-building as well as coastal and oceanic sailing vessels.
Such a massive increase in the supply of transport services, even without major improvements in technology, allowed significant convergence of prices within the Atlantic economy. Thus, as economists McCloskey & Zecker point out (pdf), wheat prices expressed in silver within the Atlantic economy narrowed from a range of 6.66:1 around 1400 to a range of 1.88:1 around 1750.
The world economy was still not globalised, was not in an era of mass trade: but the Atlantic economy came to be somewhat so. In such a situation, the price-specie-flow mechanism increasingly became irrelevant within the Atlantic economy–specie prices adjusted within what was effectively a common specie market rather than quantities shifted between distinct specie markets. That is, there came to be something close to common gold and silver prices in the Atlantic economy. It is striking that, as economist David Glasner notes in this comment (links added):
... it seems that Adam Smith, David Hume’s very good friend, seems to have rejected PSFM even though in his Lectures delivered about a decade before the Wealth of Nations  was published, he gave an accurate rendition of PSFM, but completely ignored PSFM in the Wealth of Nations. So even in Hume’s time, it may be questioned whether PSFM was the right theory.
I would say, PSFM was the right theory for the global trading networks (though that would stop being true from the 1820s onwards) but already the wrong theory for the Atlantic economy (or any mass trade system). This being a major indicator of why one can talk of a transition period before genuine economic globalisation because of the depth of trade interactions in the Atlantic economy.
Once steam technology starts driving down transport costs (on sea and and on land) and the telegraph drives down communication costs (ditto), then we are in the world of mass trade (i.e. globalisation). Including the politics of mass trade and so of globalisation.
Trade in goods becomes sufficiently large, and prices sufficiently converging, that trade began to seriously affect general factor of production (land, labour, capital) incomes. Creating the politics of globalisation: specifically, mass politics of mass trade (pdf). Not the patronage politics of network trade one had had earlier, with its licences and monopolies.
With the scale of trade that economic globalisation entailed, scarce factors of production attempted to restrict in the inflow of goods in order to protect their scarcity premium. Plentiful factors of production sought access to global markets to broaden their income sources (i.e. reduce their plenty-penalty).
How could one tell if a factor of production was scarce or not? If you imported it, it was scarce; if you exported it, it was plentiful. In the case of land the indicator was whether you exported (plentiful) or imported (scarce) its products. Thus, Britain exported labour and capital; hence labour and capital allied to force free trade on scarce land. Germany exported labour and imported capital; so scarce land and labour [capital] allied to force protection on plentiful labour. The settler societies (such as the US and the Antipodes) imported labour and capital; so labour and capital allied to force protection on plentiful land.
As modern economies became services-dominated economies, the above effects became increasingly muted, since participants in such (largely not internationally traded) industries benefited from access to cheaper goods from access to world markets without (usually) being threatened by imports from the same.
The period of globalisation had its own sub-periods in terms of the general barriers to trade (pre 1914, 1914-1945, post 1945) and wider patterns. The wider patterns being whether international trade was dominated by mass commodities following generalised comparative advantage, or by more narrow, specific goods and industries patterns due to economies of scale. To put that another way, shifts between production and trade reflecting generalised geography versus path-dependent, geographically much narrower, specialisation of production.
Such a shift splintered factor of production effects on income from expanded trade. That, along with the expansion of the importance of services in developed economies, encouraged the retreat of trade protection and expansion of free trade.
If you want to understand the patterns of trade in a globalised world, I recommend listening to Paul Krugman’s 2008 Nobel memorial prize speech. (In fact, if you are interested at all in how to do social science research in general, I recommend listening to the speech here.)
Network trade versus mass trade
It is a great mistake to look back on the history of trade without grasping how very different pre-globalised trade is from trade in a globalised economy; how different network trade is from mass trade. With the partial exception of the Atlantic economy noted above, pre-globalised long-distance trade was a matter of networks. These networks ran through and between localised farming economies whose activities were only minimally affected by said trade networks. (Pastoralist economies were more deeply penetrated by trade, but they were also much smaller in population, much more dispersed, with lower transport costs.)
Trade mattered, however, a great deal to states, because revenue from trade was relatively easy to access (involving, as it did, foreigners and specific nodes and routes) and had positive economies of scale (that is, tended to go up faster than the expansion in territory controlled). So the scale (both extensively in their territorial extent and intensively in penetration of territory controlled) of agrarian states was strongly affected by levels of trade. Hence the mass collapse and decline of states when the Roman Empire’s decline in silver production led to a major reduction in Eurasian trade networks in the early C3rd.
Trade mattered to states, state elites and (if the state forbore sufficiently from expropriation) non-state elites. It was not an engine of broad economic growth: as there essentially was no such thing before the Industrial Revolution (with the possible exceptions of early modern NW Europe, Yangtze River valley and Tokugawa Japan). Even in those cases, trade was not a driver of growth: it simply was not a big enough factor to be so. (Or, for that matter, the right sort of factor.)
Localised, not integrated
But that [importance for states and elites] is not to be confused with societies being deeply penetrated by long distance trade. Agrarian societies were overwhelmingly localised farming societies where most people never ventured beyond local areas. The larger the area considered, the more dubious is seeing the localised farming communities therein as constituting a single, coherent “society”. Precisely because production was overwhelmingly so local (in production and consumption) and movement by people so limited, religion and language were the main connecting forces, not commerce.
The cycles of agrarian life and production were common across localities in the sense of being repeated in many localities, but were not common in the sense of connecting across localities. Religion encouraged convergence by providing framings and identities not bound by localities: particularly if it had strong institutional structure (the Catholic or Orthodox Churches) or strong social role (Islamic scholars with Sharia, Brahmins with Manusmrti).
States themselves were, at best, limited integrating agents because it was entirely possible to have states whose rulers and agents did not come from where they ruled while said rulers and agents dominated (or entirely monopolised) political decision-making. Islamic and Hindu states were particularly weak integrating agents as provision of law was dominated by clerics, not rulers. Trade was even less of an integrating agent because it relied not at all on converging of said localised economies in some larger coherent society and, even during periods of heightened trade, was a small part of total economic activity regardless of how important it was to states and elites.
Being a denizen of contemporary societies–highly urbanised societies with mass literacy, massive levels of international trade and pervasive communication with strongly integrating states–is to live in a very different social environment. The, largely unthinking, assumptions that come from living in such societies are a very poor guide to understanding the social realities of highly localised farming societies in the pre-globalised world of thin trade networks.
The Industrial Revolution, once it gets seriously underway from the 1820s onwards, is a transformative experience in so many ways. It takes considerable effort to not bring its profoundly different perspectives to the study of the past.
[Cross-posted from Thinking-Out-Aloud.]