The gate that did the work is a deposit balance rather than a position limit, so it filters by account size and nothing else: an individual now has to show 30 million won, which at the 1,380.9 won to the dollar implied by the exchange's own conversion of the QLD flow is roughly 21,725 dollars, where 10 million won was about 7,242 [1][14]. Tripling it in one step reassigned the product to accounts of a different size.
The flow arithmetic suggests the accounts that left were traders rather than holders. Between the simultaneous listing of all sixteen products on May 27 and July 30, individuals bought a net 15.2876 trillion won of them, according to Korea Exchange figures reported by Sedaily [4], while combined net assets moved from 5.0075 trillion at listing to 5.8534 trillion, a gain of 845.9 billion, or one won of net asset growth for every eighteen won of net retail buying [8][18]. Something absorbed the difference, and the exchange data as supplied does not separate liquidity providers selling down inventory from value lost inside leveraged wrappers; either way the money was passing through rather than settling. The exit has the same shape. The 1.773 trillion won of net selling is 11.6% of what had gone in [13], and the SK hynix products handed back 12.5% of their inflow against 9.9% for the Samsung ones [19].
Calling the residue an offshore leak overstates it, or rather, overstates the part that crossed a border. The 258.7 billion won that went into ProShares Ultra QQQ between the 3rd and the 28th is 14.6% of the money that left the sixteen products [16], while the domestic substitutes are far larger: KODEX Leverage, KODEX KOSDAQ150 Leverage and KODEX 200 Futures Inverse 2X together turned over 3.5014 trillion won a day in the same window, more than three times the 1.059 trillion the sixteen managed [17].
So, a view I would drop quickly on September numbers: the rule found the marginal account instead of the marginal trade, and the appetite it displaced went mostly into index leverage, which is where a regulator worried about two names amplifying the whole market would rather have it sit [10]. The counter-thesis is that a deposit test is a wealth test that says nothing about the risk being taken, and that the same doubled exposure is available in a US-listed fund Korean retail bought sixth-most heavily in the period [12], which makes the measure lost domestic turnover rather than lost risk. What would break my reading is net assets resuming growth from 5.8534 trillion won while turnover stays near a trillion a day [8], because that would mean the sixteen products kept their owners and lost only their churn, and the 1.773 trillion was a liquidity event rather than a demand one.
