The Size Distribution Of Metaorders

A metaorder is the full parent order an institution decides to execute — say, buy 2 million shares over the day — which then gets sliced into many small child orders sent to the market over minutes or hours. When researchers reconstruct metaorder sizes from broker or exchange data, the distribution of those sizes turns out to be heavy-tailed, commonly modeled as roughly a power law: most metaorders are modest, but a small fraction are enormous, and that tail carries a disproportionate share of total traded volume.

This matters directly for market-impact modeling, because the standard finding that impact scales roughly with the square root of order size is fit and validated against exactly this heavy-tailed population — a handful of very large metaorders provide most of the identifying variation in the largest-size region of any impact curve, so the fit there is only as reliable as how well that tail is captured. It also matters for detecting institutional activity from the tape: if metaorder sizes were roughly normal, "big trade today" would just mean unusually large but ordinary; because the distribution is heavy-tailed, a truly enormous detected metaorder is a meaningfully different, rarer event than the typical large one, not just a bigger version of the same thing.