This is a recurring pattern in imperial China's population records, albeit here in extreme form. The returns (administrative tallies rather than what we think of as a full census) may be telling us about more than just population movement, rather they may reveal the condition of the bureaucratic recording apparatus or the state of dynastic power, or the extent of disruption (both administratively and to everyday life) during and after periods of crisis; and sometimes they may indicate what was important to those doing the counting.

I don't have the 280 return to hand except in Deng's rendition, but his 2.99 million households and population of 18.46m certainly imply a disastrous post-Han plunge. However the following three centuries of near-stagnation followed by a near-tripling in under a century and (after another nosedive) another tripling in barely a century under the Tang (followed a few centuries later by yet another under the Song) should alone give us cause to query the numbers.

It's usually easier to give something than to take it back, but with population it's easier to take it away than put it back. And while doing away in a few generations with two-thirds of the people of one of the world's two great empires of the day might be considered a pretty tall order, tripling those left in the space of a century is still less likely, requiring a sustained annual growth rate in excess of 1% in an age when growth of 0.1% was more normal, 0.2% a notable accomplishment, 0.3% the preserve of high-performing outliers and anything more a manifestation of wholesale expansion into abundant virgin land which in this period wasn't happening on anything like the required scale.

The conclusion has to be that the returns of the 3rd-6th centuries (and indeed most of those down to the 11th) don't just tell us China's population: rather they indicate either limitations to dynastic/bureaucratic capacity reflected in deficient data-gathering or compilation, or perhaps a shift in official practice that simply treated most households as not worth recording – perhaps alongside relief from associated taxation whether as a stimulus to growth or again out of eroded government capacity.

Comparing with later cycles and given the apparent absence of growth until Sui times and the constrained growth rates characteristic of the period, I'm inclined to concur with McEvedy that something above 40m is (leaving aside short-term ups and downs) a good deal more plausible then the catastrophic falls and implausibly rapid recoveries indicated by official tallies: Tang might have started from a similar level, Song from a substantially higher one if we are to accept later realistic household numbers.

Useful long-run official series (and the contending positions in the population debate) are provided by Kent Deng, Fact or fiction? Re-examination of Chinese premodern population statistics, LSE economics working paper 76/03 (Jul 2003), and John D Durand, The population of China AD 2-1953, Population Studies 13:3 (Mar 1960). Other renditions of the long-run trend (e.g. McEvedy & Jones, Atlas of world population history, Penguin 1978; Angus Maddison, Chinese economic performance in the long run, OECD 2007) essentially follow in the Durand tradition, though Deng's spirited defence of the imperial returns is echoed in widespread acceptance of the later tallies.