This paper aims to shed light on why the downturn in global trade during the intensification of the financial crisis in 2008Q4–2009Q1 was so severe and synchronised across the world and also examines the subsequent recovery in global trade during 2009Q2–2010Q1. The paper finds that a structural imports function which captures the different and time-varying import-intensities of the components of total final expenditure – consumption, investment, government expenditure, exports, etc. – can explain the sharp decline in global imports of goods and services. By contrast, a specification based on aggregate total expenditure cannot fully capture the global trade downturn. In particular, panel estimates for a large number of OECD countries based on the individual components of expenditure suggest that the high import-intensity of exports at the country-level can explain a significant proportion of the decline in world imports during the crisis, while declines in the highly import-intensive expenditure category of investment also contributed to the remaining fall in global trade. At the same time, the high and rising import-intensity of exports also reflects and captures the rapid growth in ‘vertical specialisation’, suggesting that widespread global production chains may have amplified the downturn in world trade and partly explains its high degree of synchronisation across the globe. In addition, the estimates find that stockbuilding, business confidence and credit conditions also played a role in the global trade downturn. Meanwhile, the global trade recovery (2009Q2–2010Q1) can only be partially explained by differential elasticities for the components of demand (although the results confirm that the upturn in OECD imports was also driven by strong export growth and the reactivation of global production chains, as well as the recovery in stockbuilding and the fiscal stimulus). This may be in part because of the many policy measures that were implemented to boost global trade at that time and which cannot be captured by the specification. The paper is also a pseudo-real-time robustness test of the specification in that the first analysis of the global trade downturn is based on the data available at the time (i.e. October 2009 vintage), while an updated analysis of the global downturn as well as the trade upturn is based on a more recent dataset (i.e. October 2010 vintage). The results for the global downturn remain robust regardless of which vintage of the dataset is used.