If you look at the past seven days you see 7 sales.

If you look at the past 14 days you see 108 sales.

If you look at the past 30 days you see 558 sales.

If you look at the past 60 days you see 1721 sales.

Most recent data would suggest that sales have fallen off a cliff. That seems very unlikely. What is more likely is that short term sales are being continually revised. As such, the short term numbers do not provide an accurate picture of what is taking place in the market.

Don't know if the "contracts signed" tool on urbandigs is more accurate.

I've been thinking about this. If the count is accurate, the market is really fried. I don't think it's possible. I believe UD uses the Streeteasy count. Is there any other way of checking sales? Calling UD for an answer.

Right, continually revised.

SE is getting the recorded-sales data from the city. The city scans and data-enters sales in clumps, so sales dated, e.g., 10/20 will hit ACRIS and SE over the course of many days. Sales aren't always reported to the city in a timely manner, either. That's the way it appears to work, anyway.

Right now the newest co-op sale is dated 10/21, but if you check tomorrow there might be more 10/21s and maybe some 10/22s and 10/23s.

So, if you're using any of this to gauge number of transactions, it's not much use except over the long term.

Nevertheless, the charts looked quite differently a couple of months ago.

It can take a few days to a few weeks for sales to be recorded, so the 30-day figure is pretty accurate, but the 1-day is a small fraction of actual sales. If you want to see a trend, you can track the 60-day number over time.

I'm just watching the UD inventory. Steadily growing again...
http://www.streeteasy.com/nyc/talk/discussion/15694-inventory-back-over-9500-again

sorry, meant to post the direct link...

http://www.urbandigs.com/charts.html

I wonder what is the margin for error?

Maly, what do you mean by "error"?

Well, SE data is not perfect, so even scrubbing for duplicates must leave a margin for error. I wonder if Noah has done 100% checks to get the margin of error. I occasionally find listings in my searches that have been sold, but are still listed as available or in contract months later.

10022 how do you know that big drop in inventory in the six-month chart can be entirely accounted for by increased sales?

Maybe some of it -- maybe a big chunk of it -- is due to sellers pulling their listings, and only a portion of it is due to a bump in sales activity.

Just wondering.....

Ok. I think you mean data quality or accuracy, which is not really the same thing as margin of error (which refers to sampling error).

Well, maybe think about it this way: even if the data isn't perfect, it probably has the same level of accuracy over time. If you are concerned that the quality of the data is changing over time, that's interesting (but I'm not really hearing that).

So, I wouldn't be concerned about accuracy of the data if you are comparing against historical levels, which is how most people use this data. If you are trying to do absolute arithmetic with the data, that's a different story.

I am actually more concerned about the noise because UD gives charts based on the 1-day data, which is random. This is the least interesting measure. I'd like to see 30-day data trend on contracts and sales. It would also be nice to have some idea of the accuracy of the data on listings, but that's more academic, you're right.

NWT: while I agree with you, I don't think it's quite right to explain it as the data gets recorded in "clumps". It's simply that the data recording is time lagged, and that SE uses "date of sale" (which I think is the correct thing to do) rather than "recording date". So, you will ALWAYS see very few sales in the last 7 days since almost no sales get recorded that quickly. So if a sale gets recorded 8 days after it occurs, the first time it shows up is after the "last 7 days" time period.

However, at least in Manhattan these days, from what I've seen most get recorded within 30 days. So the difference between 30 days and 60 days is meaningful, but I think you have to make an adjustment on the 30 days for the lack of recording of the most recent closings. What I guess I'd really like to see is the number of sales from 2 weeks to 6 weeks ago compared to 6 weeks to 10 weeks ago. I'm not familiar with the SE tools for insiders, so I don't know what's possible.

Yeah, I'm probably mis-guessing the cause of the ebb and flow of data. I'll do a query every day or so on the last 14 days of co-op closings, and when nothing's new I imagine piles of forms stacking up downtown.

SE does let you edit the query URL (e.g. <<.

(Another ten tries and I'll remember how to escape angle brackets.) SE does let you edit the query URL (e.g. less than 21 days rather than less than 14 days) but doesn't let you do a search like "more than 14 days AND less than 60 days."