Was messing around with the historical data tables on the MS site. Figured I'd pop some of it into Excel and see if that helps see through some of the smoke relative to where we were and where we are. Somebody here used to parse out the impact of apartment size (# bedrooms) on the aggregate MS quarterly data that always gets (mis)reported but I havent seen that in a while. Clearly the 3br and 4br+... [more]
Was messing around with the historical data tables on the MS site. Figured I'd pop some of it into Excel and see if that helps see through some of the smoke relative to where we were and where we are.
Somebody here used to parse out the impact of apartment size (# bedrooms) on the aggregate MS quarterly data that always gets (mis)reported but I havent seen that in a while. Clearly the 3br and 4br+ sales add HUGE disruptions to the averages (and who cares about those places anyway;)
So here's the Manhattan coop average price data back to 2000:
http://aggregate-data.millersamuel.com/home/query_report?annual=0&area=17&housing_type=1&measure=3®ion=1&year_end=2011&year_start=2000
I wish I could post a chart here but I HIGHLY recommend copying the table into excel and comparing the 3 curves. Quite different behavior in the studio, 1br and 2br data sets -- slow march down from 2008 for studios and 1brs but BIG correction in 2br then a quick march UP since 2009. These submarkets are very very different. BUT, in each case they seem to be off peak now by about the same 20-22% depending upon where you think peak was. See for yourself - worth the effort!
Now the question - what are the shortcomings in the MS data? timing? validity? lag? redundancies? etc etc What holes can be poked in this?
I did the charts using median also and the studio and 1br curves are pretty much identical. The 2br median curve is more peaky but the net movement is similar.
(I have a theory now on exactly WHO in NYC the bank bailouts really helped based on these charts, but that's another discussion)
[less]
Response by somewhereelse
over 14 years ago
Posts: 7435
Member since: Oct 2009
"Somebody here used to parse out the impact of apartment size (# bedrooms) on the aggregate MS quarterly data that always gets (mis)reported but I havent seen that in a while. "
I'm just waiting on the new data. Also note, I do it with medians, not averages...\
Lag is definitely an issue...
Ignored comment.
Unhide
Response by West34
over 14 years ago
Posts: 1040
Member since: Mar 2009
Sorry SWE for forgetting it was you. Your analysis was always much appreciated.
Ignored comment.
Unhide
Response by JonathanMiller
over 14 years ago
Posts: 20
Member since: Sep 2008
West34: Glad that you are able to use the data I have parsed out for the public on my site. Since you wanted to know its shortcomings, I'd thought I'd weigh in on my own work: LAG: this is based on closed data so it is subject to the lag between contract and closing date which is 45-60 days on average. VALIDITY: using all the data accessible at the time of the report release (usually 1 business day after the quarter ends). A weakness for validity could be the fact that if I reported on the market 6-12 months later, I'd have even more sales - the trade off would be the lack of use. I've reverse tested the data so that when I go back to an earlier period and fill in the recorded data not yet recorded, the patterns are virtually the same. Since the data that gets recorded is random (not skewed towards an apartment size) I believe the info we publish is valid. REDUNDANCIES: The data tables on my site have the luxury of being more granular since there is unlimited space to present it than the reports and a lot of the regional data overlaps to allow you to parse things out in different ways. The total of all the data on the site is higher than the actual activity since one transaction in Carnegie Hill would also be included in the Upper East Side and East Side and Manhattan, etc. TIMING: I feel that's OUR best attribute since we release the report and upload the data on or around the first business day of the month following the end of the quarter.
somewhereelse - please clarify the misreported part - do you mean misinterpreted? As in the media misreads what I report? On your lag comment, if you are referring to closed sales versus contracts, I addressed that above. If that's not what you mean - would love to understand what you are referring to.
Ignored comment.
Unhide
Response by somewhereelse
over 14 years ago
Posts: 7435
Member since: Oct 2009
JM, I was just quoting W34 on that line... that being said, I do think the overall medians have always suffered from composition bias...
Ignored comment.
Unhide
Response by West34
over 14 years ago
Posts: 1040
Member since: Mar 2009
JM -- Thanks! The snarky (mis)reported comment was originally mine. It refers to the media's general inclination to cherry pick rosy results for headlines while the actual dynamics based on sub-market data are more complex (and frequently mixed).
Ignored comment.
Unhide
Response by JonathanMiller
over 14 years ago
Posts: 20
Member since: Sep 2008
Gotcha to you both - thanks. One comment though on "composition bias" - I think you really mean skew. Bias infers a non-random selection of data. I think you are referring to this: it is not uncommon for all the sub categories (no. of bedrooms) to move in the opposite direction of the overall number because of a shift in the mix of sales to larger or smaller units. I have the commentary in the report speak to these variances when they exist -otherwise I am merely a robot regurgitating the top level numbers. Make sense? If composition bias is the correct phrase for the explanation I gave, then I agree completely.
Ignored comment.
Unhide
Response by bjw2103
over 14 years ago
Posts: 6236
Member since: Jul 2007
I think anyone spending just a little bit of time with the data would agree that the 4+BR (and the 3BR, to a lesser extent) grouping is inherently subject to SSS. I'm not even sure why it's worth reporting on its own - what's the use in seeing so many $6m fluctuations QoQ? Even the 3BRs suffer from this - up over $1m in the past quarter. As West34 did, parsing out the studio, 1BR, 2BR, and overall medians will give you a fairly comprehensive story on what happened. Even if I were looking at the 3+BR market, I'd rather pay attention to those trends.
Ignored comment.
Unhide
Response by West34
over 14 years ago
Posts: 1040
Member since: Mar 2009
JM - yes, skew! We've discussed the issue here and SWE has been our skew awareness champion. Any possibility you could add charts to your site? A picture is worth 1000 words and line graphs of mean prices, by quarter, by size, since 2000 would be eye-opening (and potentially viral!)
I should have said "more" charts :-) None of those address the skew factor and the notion that it's essential to track the submarkets to see what's really happening with prices. Look at the coop line here:
http://millersamuel.com/charts/gallery-view.php?ViewNode=1168394442sGxaK&Record=0 It's no wonder some are arguing that the "market has recovered" or whatnot. The submarket lines tell a VERY different story.
Ignored comment.
Unhide
Response by JonathanMiller
over 14 years ago
Posts: 20
Member since: Sep 2008
bjw2103 - yep 4± bed is thin but 3-beds are not. In fact 3-beds were one of the key characteristics of a changing market during the boom. Solid suggestion - yes revamping my online stuff this year and interactive charts are on the wish list.
West34 - completely agree with the submarket observation. I remember when so many high end new devs were closing circa 06 and 07 that were skewing the market up higher even though most people were not seeing prices rise all that much if at all. This was reflected in my commentary at that time.
Again - thanks for the insights.
Ignored comment.
Unhide
Response by stevejhx
over 14 years ago
Posts: 12656
Member since: Feb 2008
Eddie Wilson? How's that "putting your profits back to work" working out for ya?
Ignored comment.
Unhide
Response by bjw2103
over 14 years ago
Posts: 6236
Member since: Jul 2007
JM, thanks. If the 3BR data is solid, what do you think accounts for the $1m+ jump in 3BR medians from Q1 to Q2? Maybe there's a particular high-end building that I'm not thinking of that caused such a drastic skew? It's pretty crazy.
shows seasonality..that bump up in Q3-2011 "2011 3 345000 563500 1211600 " on the higher end was the result of deals signed March-June of that year. Bids coming in today and deals signed today, will ultimately power the Q3-2012 report.
So if we r trying to use median sales data to quantify 'where bids are coming in today', just understand its a 2 qtr lag due to time it takes to go from contract signing to close + lag in ACRIS filings.
Ignored comment.
Unhide
Response by West34
over 13 years ago
Posts: 1040
Member since: Mar 2009
I picture you coughing into your hand while muttering "lagging data" whenever JM walks by at industry events ;)
Ignored comment.
Unhide
Response by urbandigs
over 13 years ago
Posts: 3629
Member since: Jan 2006
no not at all..JMs well aware of the lag, the problem is the sales data is the only verifyable dataset we can count on for price discovery. We both just wish the city could speed up the process and work on getting complete info public in a much shorter time span.
I just want the industry to recognize the lag, and know what past marketplace these #s are confirming.
'/%3e%3c/g%3e%3c/svg%3e)
"Somebody here used to parse out the impact of apartment size (# bedrooms) on the aggregate MS quarterly data that always gets (mis)reported but I havent seen that in a while. "
I'm just waiting on the new data. Also note, I do it with medians, not averages...\
Lag is definitely an issue...
Sorry SWE for forgetting it was you. Your analysis was always much appreciated.
West34: Glad that you are able to use the data I have parsed out for the public on my site. Since you wanted to know its shortcomings, I'd thought I'd weigh in on my own work: LAG: this is based on closed data so it is subject to the lag between contract and closing date which is 45-60 days on average. VALIDITY: using all the data accessible at the time of the report release (usually 1 business day after the quarter ends). A weakness for validity could be the fact that if I reported on the market 6-12 months later, I'd have even more sales - the trade off would be the lack of use. I've reverse tested the data so that when I go back to an earlier period and fill in the recorded data not yet recorded, the patterns are virtually the same. Since the data that gets recorded is random (not skewed towards an apartment size) I believe the info we publish is valid. REDUNDANCIES: The data tables on my site have the luxury of being more granular since there is unlimited space to present it than the reports and a lot of the regional data overlaps to allow you to parse things out in different ways. The total of all the data on the site is higher than the actual activity since one transaction in Carnegie Hill would also be included in the Upper East Side and East Side and Manhattan, etc. TIMING: I feel that's OUR best attribute since we release the report and upload the data on or around the first business day of the month following the end of the quarter.
somewhereelse - please clarify the misreported part - do you mean misinterpreted? As in the media misreads what I report? On your lag comment, if you are referring to closed sales versus contracts, I addressed that above. If that's not what you mean - would love to understand what you are referring to.
JM, I was just quoting W34 on that line... that being said, I do think the overall medians have always suffered from composition bias...
JM -- Thanks! The snarky (mis)reported comment was originally mine. It refers to the media's general inclination to cherry pick rosy results for headlines while the actual dynamics based on sub-market data are more complex (and frequently mixed).
Gotcha to you both - thanks. One comment though on "composition bias" - I think you really mean skew. Bias infers a non-random selection of data. I think you are referring to this: it is not uncommon for all the sub categories (no. of bedrooms) to move in the opposite direction of the overall number because of a shift in the mix of sales to larger or smaller units. I have the commentary in the report speak to these variances when they exist -otherwise I am merely a robot regurgitating the top level numbers. Make sense? If composition bias is the correct phrase for the explanation I gave, then I agree completely.
I think anyone spending just a little bit of time with the data would agree that the 4+BR (and the 3BR, to a lesser extent) grouping is inherently subject to SSS. I'm not even sure why it's worth reporting on its own - what's the use in seeing so many $6m fluctuations QoQ? Even the 3BRs suffer from this - up over $1m in the past quarter. As West34 did, parsing out the studio, 1BR, 2BR, and overall medians will give you a fairly comprehensive story on what happened. Even if I were looking at the 3+BR market, I'd rather pay attention to those trends.
JM - yes, skew! We've discussed the issue here and SWE has been our skew awareness champion. Any possibility you could add charts to your site? A picture is worth 1000 words and line graphs of mean prices, by quarter, by size, since 2000 would be eye-opening (and potentially viral!)
West34, see here:
http://millersamuel.com/charts/
I should have said "more" charts :-) None of those address the skew factor and the notion that it's essential to track the submarkets to see what's really happening with prices. Look at the coop line here:
http://millersamuel.com/charts/gallery-view.php?ViewNode=1168394442sGxaK&Record=0
It's no wonder some are arguing that the "market has recovered" or whatnot. The submarket lines tell a VERY different story.
bjw2103 - yep 4± bed is thin but 3-beds are not. In fact 3-beds were one of the key characteristics of a changing market during the boom. Solid suggestion - yes revamping my online stuff this year and interactive charts are on the wish list.
West34 - completely agree with the submarket observation. I remember when so many high end new devs were closing circa 06 and 07 that were skewing the market up higher even though most people were not seeing prices rise all that much if at all. This was reflected in my commentary at that time.
Again - thanks for the insights.
Eddie Wilson? How's that "putting your profits back to work" working out for ya?
JM, thanks. If the 3BR data is solid, what do you think accounts for the $1m+ jump in 3BR medians from Q1 to Q2? Maybe there's a particular high-end building that I'm not thinking of that caused such a drastic skew? It's pretty crazy.
Steve, wrong thread.
JM -- when is the 1st Qtr 2012 data going to be here?:
http://aggregate-data.millersamuel.com/
Thru 1st Qtr 2012 Miller Samuel data parsed by apartment size. Conclusion: FLAT MARKET!
Median Manhattan coop prices:
QTR Studio 1 BR 2 BR
2012 1 360000 546000 1045000
2011 4 345000 555000 1050000
2011 3 345000 563500 1211600
2011 2 355000 550000 1160000
2011 1 347500 535000 995000
2010 4 345000 568000 1150000
2010 3 350000 550000 1175000
2010 2 380000 599000 1100000
2010 1 330000 575000 1375000
2009 4 350000 591000 1130000
2009 3 367500 575000 1450000
2009 2 365000 580000 1750000
2009 1 365000 580000 1125000
2008 4 382500 645000 1450000
2008 3 400000 630000 1351000
2008 2 440000 695000 1412250
2008 1 465000 690000 1417000
2007 4 415000 682000 1350000
2007 3 352000 595000 1295000
2007 2 367500 600000 1185000
2007 1 345000 549000 1120000
2006 4 385000 582500 1490000
2006 3 357500 585000 1100000
2006 2 370000 575000 1265000
2006 1 383000 612000 1217300
2005 4 349000 575000 1195000
2005 3 352000 579000 1100000
2005 2 327500 525000 1300000
2005 1 295000 500000 915000
2004 4 285000 455000 879000
2004 3 275000 446000 875000
2004 2 247000 410000 799000
2004 1 237000 395000 825000
2003 4 235000 399000 760000
2003 3 229000 375000 735000
2003 2 225000 375000 695000
2003 1 225000 370000 685000
2002 4 225000 370000 710000
2002 3 220000 357000 715000
2002 2 205000 350000 670000
2002 1 195000 330000 625000
2001 4 199000 325000 635000
2001 3 198000 330000 652000
2001 2 187500 330000 650000
2001 1 185000 316000 650000
2000 4 165000 305000 625000
2000 3 159000 300000 655809
2000 2 147000 279000 645000
2000 1 135000 262500 575000
Number of sales:
QTR Studio 1 BR 2 BR Total
2012 1 241 609 335 1185
2011 4 196 374 283 853
2011 3 258 449 499 1206
2011 2 219 550 464 1233
2011 1 235 575 480 1290
2010 4 171 480 372 1023
2010 3 162 506 486 1154
2010 2 189 478 366 1033
2010 1 172 440 351 963
2009 4 297 532 276 1105
2009 3 242 353 259 854
2009 2 144 302 192 638
2009 1 108 165 109 382
2008 4 253 378 306 937
2008 3 333 544 395 1272
2008 2 347 424 404 1175
2008 1 313 320 319 952
2007 4 423 415 370 1208
2007 3 337 758 603 1698
2007 2 333 557 596 1486
2007 1 309 619 724 1652
2006 4 295 525 539 1359
2006 3 207 374 428 1009
2006 2 132 294 327 753
2006 1 193 352 326 871
2005 4 187 294 314 795
2005 3 230 416 393 1039
2005 2 218 431 424 1073
2005 1 219 445 461 1125
2004 4 252 436 455 1143
2004 3 316 529 524 1369
2004 2 216 460 498 1174
2004 1 239 417 474 1130
2003 4 203 639 622 1464
2003 3 283 511 686 1480
2003 2 238 514 577 1329
2003 1 221 448 484 1153
2002 4 281 478 540 1299
2002 3 320 521 611 1452
2002 2 321 592 738 1651
2002 1 192 504 556 1252
2001 4 253 495 541 1289
2001 3 206 390 605 1201
2001 2 159 228 449 836
2001 1 192 442 450 1084
2000 4 385 471 568 1424
2000 3 272 530 425 1227
2000 2 214 457 468 1139
2000 1 231 544 555 1330
shows seasonality..that bump up in Q3-2011 "2011 3 345000 563500 1211600 " on the higher end was the result of deals signed March-June of that year. Bids coming in today and deals signed today, will ultimately power the Q3-2012 report.
So if we r trying to use median sales data to quantify 'where bids are coming in today', just understand its a 2 qtr lag due to time it takes to go from contract signing to close + lag in ACRIS filings.
I picture you coughing into your hand while muttering "lagging data" whenever JM walks by at industry events ;)
no not at all..JMs well aware of the lag, the problem is the sales data is the only verifyable dataset we can count on for price discovery. We both just wish the city could speed up the process and work on getting complete info public in a much shorter time span.
I just want the industry to recognize the lag, and know what past marketplace these #s are confirming.