Irvine Housing Blog

I remember feeling, back in 2000, that the market was overinflated. We’re currently rewinding to 2000 prices which, once attained, will be a plateau from which we descend to 1992 or 1995 values.

It’s not just a matter of correcting the problems which already existed in 2000 - problems due to overenthusiastic purchasing not caused by loose lending - but also adjusting to shrinking incomes in the middle class over the past decade.

Shady Canyon gets lumped in with Turtle Rock in the MLS data. We may be able to separate these transactions out by the year of construction, but I don’t think there are enough resales to make a good stand-alone analysis on it yet.

Curious,

We totally agree!  While this is the *Irvine* Housing Blog, Irvine prices are affected by prices in Tustin, Lake Forest, Costa Mesa, etc. due to a reasonable substitution effect—and it would be worthwhile to understand how a hedonic price index model is moving in those other areas and, ultimately, all of Orange County. 

Because Case-Shiller does not publish data for OC (only LA/OC), most “Orange County” price indexes are median-based.  That’s a bit of a shame.

octal77,

Good question about Shady Canyon.

In theory Shady Canyon sales should have been largely ignored under the overall model exclusions (such as no sales of over $2M and no lots over a certain size).  It’s always possible that a very small number of data points slipped in as data errors (where neighborhood was labelled as Turtle Rock or Turtle Ridge) but it should be minimal, and it would only be at the very bottom of the Shady Canyon market.

Also, thank you for pointing this out.  Good models refresh over time.  In a hedonic modeling system, it’s easy to add a new neighborhood or consolidate like areas in a forthcoming edition.  We are always looking at areas of investigation to heighten model accuracy.

octal77 & IrvineRenter - I wasn’t shocked at all by both TRs being at the top of the list _and_ their premiums.

IrvineRenter mentions that Northwood Pointe also feed into the Irvine school district but not all Irvine schools are created differently.

What is different about Turtle Rock, Turtle Ridge, and University Park?  They feed into University High school, which was recently ranked as the #8 school in the United States.  I think Northwood came in at #96 and Irvine High came in the #460s.  And who keeps track of these statistics?  Asian parents and I should know.  I was a product of University High whose parents moved to Irvine (and Turtle Rock) specifically so that my sister and I could attend University (and this was back in the open enrollment days too).  So, I think the TR premium is partly to do with the Asian influence.

Cheers.

Thank you. I have corrected the chart.

The Holy Grail would be combining the data you suggest with current financials of any troubled properties.

when you are checking redfin, is there a way to find out how much mello roos a property makes you pay?  redfin information shows if there is or isn’t mello roos, but not the percentage.

I agree about the school systems in neighbhoring cities.  I think the Irvine schools are what keeps the values of the homes higher during the education crisis.  We bought in Mission Viejo thinking SVSD would be fine…but now we have two young children that will soon be entering the elementary years and are considering private schools with the mess that SVSD is in.  I can see how buying an Irvine property would be very desirable when looking at schools—especially with the budget school cuts.  There are probably less budget cuts to Irvine because more affluent parents donate/contribute to the school.

Irvine USD is a public school district, so the majority of its funding is fed off from the state budget. IUSD does feel the pain with the budget cuts. The one difference is that I was told there are private doners to IUSD that help offset some of the budget cuts.

Hi RB,

Thanks for your insights.  In the next post (Wed Jul 13th), we are going to get into some detail about the coefficients for each explanatory factor in the model.  I’d like to save some specifics for that post, but I can respond to a few general points you’ve made.

We’ve included the same three factors as you, along with 4-5 others that added predictive power to the model.  The negative valuation of a bedroom is not uncommon in hedonic housing models from the admittedly modest research of other studies I’ve done.  The general intuition is that adding a bedroom doesn’t inherently add value unless it’s coupled with a sufficiently large increase in square footage.  For example, a 3/2 2000 sq. ft. property changing into a 4/2 2000 sq. ft. property isn’t a value add because you’ve created the bedroom at the expense of other living space.  Doing so might even destroy value if the market demands larger communal living spaces.

We are using a full 10-year sales dataset (and creating a year-quarter dummy variable) so that we can decrease the standard error / p-values in the overall model.  We’ve also run the analysis year-by-year and in year groups, and the coefficient p-values and standard errors increase as we shorten the timespan of the underlying dataset.  You are perhaps seeing p > 0.05 when using a 3-month dataset due to the size of the dataset as opposed to the formulation of your model.

We have the same basic issue:  bed-bath-sqft are all correlated to some degree.  In our data, these correlations are bed-bath 0.603, bed-sqft 0.729, and bath-sqft 0.776.  Whether this level of correlation is a problem is some matter of debate and seems to be an eternal issue among academic statisticians.  While these correlations are higher than ideal, they are not drastically altering the meaningful nature of the regression coefficients IMHO.

We used the vif() function in R’s library(car) to find that VIF (Variance Inflation Factor) is between 1.0 and 2.50 for each coefficient.  While many will have an opinion on “how high is too high” for VIF, the 1.0 to 2.50 range that we obtained seems reasonable compared to the upper bounds of 5 and 10 that are often proposed.