Kevin Boer thought he found an error in Redfin’s accounting of its MLS results. What he found turned out to be trivial, which led to another round of war-hooping from the Redfin tribe.
Meanwhile, our new contributor James Hsu has demonstrated that Redfin’s horse runs behind the middle of the pack among big-name Seattle brokerages. In other words, as predicted, experienced traditional agents do out-perform Redfin’s salaried agents.
I finally took a look at Redfin’s spreadsheet today, which they were kind enough to share with me. There are two formulae for calculating the Sales Price to List Price ratio, but I’m not sure that matters. Ten houses sold for less that 65% of list, which I find amazing. More amazing still, nine sold for more than 150% of list. One of them sold for 1,068.526% of list.
One condominium sold for 10% of list price. At that price, I think I might have taken more than one. Condo buyers are smarter, though. Only four of them were willing to pay more than 144%, although a whole bunch sold for more than 110% of list. In Phoenix, they’d be investigating for loan fraud.
Here’s the cute part: Redfin sold 45 condominiums, of which 20 sold for more than its vaunted average performance of 99.340%. Okayfine, fewer than half. For residential listings, however, Redfin kindasorta sucked: Out of 125 sales, 70 homes sold for more than their average.
I named all kinds of reasons for holding Redfin’s claims in doubt. The overarching question — tough agents or tough clients? — is the one Redfin seeks to avoid. Its claims all week have been a textbook example of the Fallacy of Affirming the Consequent: If P then Q, Q therefore P. If Redfin’s agents are tougher than average, then its ratios should beat the market. Redfin’s ratios beat the market (a specious if not actually false claim in any case), therefore Redfin’s agents are tougher than average. The conclusion does not follow, and the raw numbers seem to argue eloquently that the results achieved by Redfin’s clients were caused by Redfin’s clients, not by its agents. The skinflints did well, and the rest did worse than the market as a whole.
Yet again, there is a way to find out for sure. In the RE.net, Glenn Kelman does not control the debate. He cites Kevin Boer, ignoring James Hsu, but all of the rest of us can see a church by daylight. This whole debate is easily settled, and I might yet be silenced utterly: Your numbers are bogus, Mr. Kelman. Show me the files…
Further notice: In my original post about this imbroglio, I cited a weblog entry by Marlow Harris that in turn cited a report by the NWMLS system. I had doubted the veracity of the claims in that report, since it asserted a much greater SP/LP ratio than I would expect. I was not endorsing the report, I was simply making the grand rounds of links that I knew about at that time. It may be that that report has since been revised or retracted — not that this has anything at all to do with any argument I have made here or anywhere. For future reference, if you want to know if I am endorsing or merely citing a link, look for language like this: “Hot damn! Katie Kastensmith is really onto something!” I know it can be hard to peer through the murky mists of my impenetrable self-restraint, but you’ll manage somehow.
Technorati Tags: blogging, disintermediation, real estate, real estate marketing
Greg Swann says:
Would someone in Seattle be gracious enough to look up NWMLS #26038601? Either the speadsheet Redfin shared with me (and others) is bogus, or a Redfin buyer paid 122.909% of list price for a home. I’d be grateful to know which is true.
March 2, 2007 — 5:41 pm
James Hsu says:
Since it’s public record, ..it did indeed sell for 122.91% of list price. Listed at 375,888 and sold for 462,000. Talked to the agent briefly just now and it apparently was in a pretty massive bidding war. Wasn’t too spectacular of a house. Split-level, built in 1967, …minimal updating. So, …it would appear that Redfin was top notch in blowing everyone out of the water with their bid. ….nevermind the fact that they got a bigger commission because of it.
March 2, 2007 — 5:53 pm
Kevin Boer says:
Ok, I think this is the last bit of energy I’ll put into this debate, but I’ve just posted another look at the data using strictly the Redfin methodology and the Redfin data. Under those conditions, their results are correct — even when excluding the egregious MLS data entry errors — and in fact stack up well against the other big firms.
James’s results are quite different from mine, but he uses median, instead of average, looks at a broader geography, and excludes condos.
Ultimately this is all a fun, but pretty meaningless, exercise. Tell me what you want me to prove, and I’ll find the data set and methodology to prove it.
March 2, 2007 — 7:56 pm
Drew Nichols says:
Kevin’s point is the best: the numbers can be manipulated various ways to find anything you want to know. Salesmen know this.
The most logical conclusion is that Redfin attracts more savvy (maybe?) buyers who are also clearly budget conscious over other factors. “Full service” clients will continue to value other factors beyond simply bottom lines. Either way, I’m with Kevin.
You can prove anything with data and methodology. Doesn’t everyone know that 80% of statistics are incorrect? 🙂
March 2, 2007 — 11:05 pm
Greg Swann says:
> Doesn’t everyone know that 80% of statistics are incorrect?
47% of all statistics are made up on the spot.
People will say, “Really?!”
This may be the most perfect joke about modernity.
March 2, 2007 — 11:14 pm
Kevin Boer says:
So this statistician goes to the firing range for a bit of practice. His first shot is one foot to the left of the bullseye. His second is one foot to the right.
“Yippee,” he yells! “On average, I hit the bullseye!”
March 3, 2007 — 12:04 pm
jf.sellsius says:
Redfin’s stats are much ado about nothing. Even if the company “average” sale-to-list price ratios were accurate, who would hire an agent based on them? Only the same folk who would bid on a zestimated house because zillows’s median error rate is 7.5% It’s like buying a gas guzzling Caddie because GM’s car line averages 30 mpg.
IMO, the more meaningful stat is customer satisfaction, though that is always suspect as well.
March 3, 2007 — 3:07 pm
Norm Fisher says:
Statistics are like a bikini. What they reveal is suggestive, but invariably, the most interesting facts remain hidden. 🙂
March 3, 2007 — 7:32 pm