The video links below were disabled for direct embedding on Youtube. Just click on the links and enjoy.
Snow Version
Colors / Dance — George Winston
Water version
Colors / Dance — George Winston
Our contributing arithmetician, Zileas, has provided an analysis of Aliso Viejo’s market for your study.
Conclusions:
– Over the last 9 months, the typical 2 or 3 bedroom condo of 1000-1400 sqft in Aliso Viejo dropped roughly $123 in value per day. I personally think that this number has gotten worse, but I can only prove $123 a day with the data I have. This comes out to about a 1% drop in value per month for the condos I looked at. This is less than half the rate of decrease in Ladera, which you would expect since Aliso is closer to the beach, is more developed, and anecdotally has fewer toxic mortgages and what not.
– For every $1 you overprice your home, you LOSE $0.81 off the final sales price… So if your home is worth $400k, two choices you might consider are (based upon the data): price at $400, sell at $400, or… Price at $425, sell at $380K. This is because buyers tend to ignore overpriced houses in this market (which has so many choices!), and move on to someone who is showing they are willing to sell. Again, sellers: Price to sell!
– If you are going to buy a 2 bedroom condo, you may as well go high on the square feet and get a bigger living room or bedroom or whatever, in my opinion. The cost of buying a condo with a bigger living space (but the same # of beds/baths/basic features) is only $88 per extra square foot in Aliso, which makes those 1400 square foot 2 bedrooms seem a lot more attractive than the 1000 sqft ones… I’d sure pay 35K for 40% more living space!
– Older condos sell for less… About $1700-$2000 less for every year older they are. Note that homes get older as you own them…
– Compared to typical AV properties similar to them, Windflower condos are worth a bit more (about 20k, which is roughly 2%). Other developments may be worth more or less than the average development, but I couldn’t prove it with the data I had.
– I’d really love wider and better data. If any Realtors want to hook me up for my benefit, and for helping you persuade your sellers to price reasonably, look me up!
What Properties Did I Look At?
I used 77 condos sold in the last 9 months in Aliso Viejo, with 2-3 beds, and 1000-1400 square feet. The properties ranged in age from 6 to 18 years old. I performed this analysis on Nov 22nd, on data that is a week old.
How Did I Value Them?
I used statistical regression, which is a fancy way of drawing a line through a bunch of points, but instead of lines which only allow you to study one thing in relation to another, it can study a lot of things in relation to one thing all together. This is the same type of technique that they use in science, pharmaceuticals, and economic models. There are math notes on the specifics for the curious or mathematical inclined at the bottom.
I valued homes using several slightly different methods and selections of properties to make sure my conclusions were robust. The key factors that went into housing value were: when it sold, when it was built, # of square feet, # of bathrooms, # of bedrooms, and how much the seller overpriced/underpriced the property (final asking price – sold price).
How do you Interpret This?
– Use it to look at properties very similar to the condos I used – it would be lousy for looking at a 5 bedroom house in Aliso Viejo, or a condo in Newport Beach.
– It is most accurate right now. This sort of analysis can get stale, things change!
– DON’T use it as investment advice, I’m not a certified professional, just a guy with advanced statistics training, an MBA, and some spare time.
– I have included some valuations of “typical” properties below. A “typical” property has average view, has average upgrades/condition, has average selling circumstances, and so on EXCEPT for the specific details I lay out. You would have to judge if a property would command an “above average” price (e.g. diamond-inlaid bathroom mirror) or a “below average” price (mold growing on walls, neighbor owns meth lab, etc) and adjust accordingly.
2 Bed, 2 Bath, 1000 sqft, built 1990, closing dec 31: $379,000 +/- 25K for relative quality of property.
2 bed, 2.5 Bath, 1400 sqft, built 1996, closing dec 31: $433,000 +/- 20K for relative quality of property
(the +/- are a “95% confidence interval” – if 20 houses sold of each of the above types, I predict only one each would fall outside that range on average)
Where are the Graphs?
Because I am plotting several things vs price, I can’t make a graph. You can graph sqft vs price, but you can’t graph beds AND baths AND year built AND sqft vs price all at once.
How NOT to Interpret this:
– Do not read too much into the $/sqft number. When you see this figure on real estate sites, you are valuing the entire property by it’s square footage. My model considers sqft to be only a part of the overall valuation (instead of 100% of it), so the number you see is smaller.
– You may be tempted to compare this to my Ladera Ranch analysis. They use slightly different methodology and data ranges. Therefore, direct comparison requires some pretty serious knowledge of how these sorts of models work, and even then you are doing a lot of hand-waving. Treat them as stand-alone unless you have deep knowledge of this stuff.
Details of Research for the Curious or Mathematical:
Basic Method;
Linear Ordinary Least Squares regression (which is indeed a BLUE regression) of various predictors on soldprice. Boxcox proved linearity with a theta of 7 (!!). VIFs, except where noted, were generally below 2 (beds and yearsagobuilt sometimes crested over it to like 2.35ish). No crazy weighting of data points or questionable pruning, no resampling my own error, no smoking crack, this is a pretty vanilla set of regressions.
All regressions were whitewashed of heteroskedasticity. And yes, there was significant heteroskedasticity because I had no 1 bedrooms and not that many 3 bedrooms in the sample.
Dataset:
This is in the basic “layman’s” posting, and trimming is described in the chart (to either 11 months (all), 9 months, or 6 months).
Predictors and Justifications Behind Using Them:
Daysago — # of days ago the property was built, captures gradual constant pricing decreases which we know exist!
Daysagosquared – Same as above, but captures accelerating trends to a degree.
Beds – More bedrooms usually means more value, and it has regressed well in other housing studies.
Baths – Same logic as above.
Sqft – Larger houses sell for more, duh.
Yearsagobuilt – Newer homes sell for more – inherently they look better, have higher tech construction, less deferred maintenance problems, etc.
Overprice – In a buyers market, overpricing means you get less offers, which should reduce the price you get for the property… Asking too much means you artificially decrease demand away from yoru actual “willing to sell” point.
Fixed Effects – some communities are gated or have great views or better layouts or have a slightly better location or were more upgraded at initial construction by the builder. It is hard to know the specifics, but a general fixed effects model can capture some of these unobservable differences. I put condos that had 7 or more sales in the same development into a group, and all others into the omitted category. 35 out of 99 condos were in this omitted category.
Regression 1, 99 condos: In this regression, I noticed that the regression was mediocre (borderline p-values on good predictors, though same signs and all that) unless I added in daysagosquared, which crudely captured an accelerating pricing trend with time. With it, the predictor coefficients became VERY good (daysago t-stat went from 1.98 to 4.48 for example).
Unfortunately, this regression was also highly multi-collinear between daysago and daysagosquared, though they tested for P<.01% for joint significance. Nonetheless, we don’t really care what houses went for in feb and march that much, so I decided to trim the data to get around having to use daysagosquared in future regressions.
Regression 2, 99 condos: This is pretty much the same regression, but I added in fixed effects as a robustness check and also to fish and see if any developments were obviously better or worse. Not useful for answering the question we all want answered, which is, what are houses worth today, but interesting nonetheless! It seems that windflower is worth 19k more (p <.1%). Nothing else could pass the null hypothesis, so I decided that, especially since I’m cutting data which would further strain it, I may as well dump fixed effects for this regression. Alas.
Regresion 3, 77 condos (Suggested): This is the regression I’m basing the majority of my conclusions on, and I think it is the best one in terms of predicting what you’d pay RIGHT NOW. The one variable that had weaker significance, baths, is one that we know does in fact have real-world significance, so I felt comfortable leaving it in with p=10.2%. Besides, the purpose of this analysis is to track price decrease more than anything, so its not doing much harm sitting in there adding a little bit of predictive power.
Regression 4, 37 condos: I trimmed the data to the last 6 months in this regression. My model started to get unstable from lack of data at this point. Among other things, beds and baths went deep into insignificance, and their predictive power appears to have gotten sucked up by other variables, especially sqft (the strongest predictor). It’s hard to make a good comparison of this to regression 3, though the general trends predicted in 3 are also predicted here.
Regression 5, 77 condos: this is for those of you who are skeptical about the overprice variable for a variety of reasons. I encourage you to think those through carefully and consider what it would mean for the variable to have different strengths, but I included this in case you consider it invalid. The regression is reasonably useful without the variable, you just drop R^2 a lot, and reach the same conclusions on the price trend.
How ‘bout those Stata Logs
Fine fine… Here is regression #3
Linear regression Number of obs = 77
F( 6, 70) = 8.52
Prob > F = 0.0000
R-squared = 0.5250
Root MSE = 18832
——————————————————————————
| Robust
soldprice | Coef. Std. Err. t P>|t| [95% Conf. Interval]
————-+—————————————————————-
daysago | 123.4795 35.79553 3.45 0.001 52.08754 194.8714
bed | 15255.4 7203.995 2.12 0.038 887.4886 29623.32
bth | 17246.57 9697.281 1.78 0.080 -2094.052 36587.18
sqft | 88.41503 25.2406 3.50 0.001 38.07424 138.7558
yearsagobu~t | -1690.508 430.1378 -3.93 0.000 -2548.391 -832.6255
overprice | -.8145235 .2480114 -3.28 0.002 -1.309167 -.3198804
_cons | 259806.1 33764.72 7.69 0.000 192464.4 327147.7
——————————————————————————
Aliso Viejo Analysis — Link to Word Document
