Around a year ago, I switched from using Vanguard index funds to Vanguard ETFs in my IRA to manage a simple 4-5 fund portfolio.
Since these holdings exist within a retirement account, this decision was, of course, not driven by any tax arguments. Rather, I found that it was simpler to perform re-balancing, since I'm coordinating a desired global asset allocation over this account, a 401(k), and taxable account.
When I go to add new funds or re-balance the existing holdings in this account, I will typically execute limit orders at some point in the middle of the trading day (e.g., 1pm EST). (I avoid placing orders in the first/last hours of the day, as there is strong historical evidence suggestive of higher bid-ask spreads during those times.) I will typically set my limit price equal to 2-3 cents below whatever the current bid price is at the time of placing the order, and the order will execute within at most an hour (usually much sooner). (I understand that the order could potentially fail to execute, and am always willing to retry at a limit price closer to the bid price.)
I believe that in the long run, over many years of consistent application, this strategy will yield (in expectation) a greater return than just buying shares of the equivalent index fund at NAV on the days I choose to add money to the account.
My reason for believing that I will come out ahead is the following. If we approximate this portfolio's performance as stochastic process calibrated over any historical time interval of sufficient length, we will, of course, observe that its expected growth is positive - e.g., it has positive drift - as the US equity market has historically grown over time.
My order is being executed a time t+h (h>0 is the time between submitting my limit order and it executing) at a price less than the bid price observed at time t. If we assume that purchasing a share of an ETF at below the bid price at time t is proportionally cheaper than purchasing the share at a NAV computed at time t (Is this a reasonable assumption?), then I believe the return to be higher in expectation if using my strategy, stochastically proportional to the extra time in market (buying intraday versus buying at NAV at market close) and the "discount" I've gained by purchasing the share at a price that is (in expectation, given the aforementioned positive drift) lower than the day's ending NAV.
Even if my hypothesis is correct, the amount of money that we are talking about saving here is almost microscopic. Thus, this is but a mere thought experiment that I'm using in attempt to learn more about ETFs. (I'm buying broad market ETFs and only "trading" them to re-balance and purchase new shares, so I don't see this as trying to "time the market" in any way.)
- Does this line of reasoning check out, or am I missing an obvious consideration?
- Much of my argument rests on the assumption that if at time t, an ETF has bid price B_t and ask price A_t, then purchasing at some price B* < B_t is proportionally cheaper than buying a share of the analogous index fund at NAV computed at time t. Of course ETFs and their corresponding index funds can have prices that are on different nominal scales (e.g., VTI and VTSMX), but their gradients (with respect to time) should be proportional. Thus, when I say "cheaper," I am simply comparing the integrals of instruments' price changes over the time in market. Is this a valid assumption, or am I oversimplifying the pricing duality that exists between ETFs and their index fund equivalents?