I am trying to study the potential market impact of extremely large ETF orders (>20% of ADV) made over relatively short time periods (~1 hour). Conventional market impact models for equities would show a fairly sizable market impact cost with a trade this size. My question is whether very liquid ETF's (around the size of XLF, FAS etc.) warrant a different market impact model from traditional equities due to the arbitrage mechanism? My thinking is this. Since these ETF's track an underlying equity index, they have a prescribed value apart from the market value. Therefore when a very large trade (>20% of ADV) occurs within an hour and begins to shift it away from the intrinsic NAV as prescribed by the underlying index, then arbitraguers will "arb" the ETF back to a point near its fundamental value. The result of this is that the market impact of large ETF orders is in fact less than the market impact of an equally large traditional equity order. Such that while it would be highly costly to move 25% of ADV through a stock within an hour, such a trade may actually be feasible with a deeply traded and arbed ETF. Any insight into this issue, and whether my reasoning is correct would be greatly appreciated.