I don't buy it, though. If it was that certain that rates were going to go up, that would be already factored into the rate of the loan. No one would sell me a 90 day lock for 6.375% if they thought long-term interest rates would be at 6.5% in June. In fact, they probably wouldn't sell me a 30-day lock at 6.25% either, because the long-term lenders would not want to fund my loan for 30 years at 6.25% if there was an upcoming rate hike.
Initially, it didn't even make much sense to me why a rate lock for 90 days would make much of a difference at all. I mean, what difference is it for someone to loan me some money from April 2006 till April 2036, vs. June 2006 till June 2036? I think the reason here must be that the person offering me a 90-day lock makes money on the margins. Someone who is actually funding my loan is happy to sell me a 30-year loan at 6.25% today, since according to their best projections, they will make money on it. However, everyone I'm dealing with is planning to keep my loan for a very short time and then resell it. (In fact, I'm dealing with brokers, who deal with resellers, who in turn deal with people who actually fund mortgages. I think.)
So my current model is that there is a reseller, who offers me a rate that's some amount higher than what they expect they can get from someone else, say, half a percent for the sake of argument. (I have no idea what this is in reality, though I'd be surprised if it's more than 1% because there's a fair bit of competition in the mortgage industry. Just check your spam folder!) Now, according to the best predictions of the funders, they will make money on loans at 5.75%, so the re-seller offers me 6.25% today. However, there is a chance that these best predictions will change based on new information. For the funders, this is not a big risk, since they're in it for the long haul. But for the resellers, it can be night and day. If the predictions cause the mortgage rates to shift by 0.5% over the next two months — unlikely, but could happen — they go from making a bunch of money on my loan to breaking even. And if the rates go higher still, they risk losing money on me.
So of course, even if the expected value of mortgage rates over the two months is flat at 5.75%, they will want to charge me a big premium so that they can assume this risk, which is pretty high for them. For me, the risk is closer to what the funders have. I may not keep the loan over 30 years, but I'm still in it for the relatively long haul. Half a percentage point difference translates to something like $80 dollars a month extra in house payments. It's not something I'd be all that happy about, but it wouldn't make me no longer able to afford the house, either. (If it did, I wouldn't be buying it!)
My wild guess is that the premium the seller is willing to pay to guard from risk is actually higher than the premium I would be willing to pay, since their risk is in some sense higher. Therefore, whatever extra amount they charge on the 90-day lock is almost by definition not worth it for me. However, writing this post, I realized I don't know as much about mortgage finance as I thought I did, and I definitely don't know enough about risk quanitification to back up this conclusion with solid reasoning.
There is a simpler argument against the 90 day lock, though. Getting a 90 day lock is effectively insurance against the long-term mortgage rates going up over the next 90 days. As I outlined in the last paragraph, increased rates are a cost that would be unpleasant, but one that I'd be able to bear. And there's a general principle that says that you should never buy insurance for something you can afford. ("Extended warranty? How can I lose?") So no 90-day lock for us.