Hedonic Modeling Mats Wilhelmsson Center for Banking and Finance (Cefin)

Hedonic Modeling

Mats WilhelmssonCenter for Banking and Finance (Cefin)

Lecture

• Tuesday 15/2– Wilhelmsson, Mats (2008). House Price Depreciation

Rates and Level of Maintenance. Journal of Housing Economics. Vol.17(1), 88-101.

– Kryvobokow, Mark and Wilhelmsson, Mats (2007). Analyzing location attributes with hedonic model for apartment prices in Donetsk, Ukraine. International Journal of Strategic Property Management, Vol.11, 157-178.

– Song, Han-Suck and Wilhelmsson, Mats (2010). Improved price index for condominums.

3

Price Variation in Space• Why do housing prices vary in space?

– That is to say, what can explain the price variation when we are using cross-sectional data?

• Attributes– Property specific

• Size, quality, age– Neighborhood specific

• Positive and negative externalities• Segmented market

• The relationship between housing price and housing attributes is estimated with the so-called hedonic regression

4

The Hedonic Regression

• Relating price to attributes of the goods• Housing, cars, electronics, wine….

• The hedonic regression is based on the hedonic value model where the property value is a function of property attributes.

• Hedonic regression controls for differences across individual properties by modeling the value effects of those differences.

• That is, we estimate an implicit (hedonic) price of the attribute– Haas (1922), Court (1937) and Rosen (1974)

First Hedonic Theory

5

The Hedonic Price Equation

• Attributes (Z)– Property specific attributes (F)

• Apartment specific attributes– Neighborhood specific attributes (O)– Time specific attributes (T)

TOFZP 321)(Price

6

The Hedonic Methodology

iy

ziz

i

i

MU

MU

yu

zup

z

P

FOC

zPyts

i

i

)(PI ..

)zu(y,max

y

i

Estimated parameters in the hedonic regression is equal to the marginal willingness to pay, that is, the hedonic price is equal to how much the individual is willing to give up of other goods to get attribute z.

7

The Hedonic Methodology

• The first stage: Estimation of P(Z) • The second stage:

• Estimation of:– Price elasticity– Income elasticity

,...),( IPgZizi

z

Pz

?

?

8

When it is used?

• Valuation/appraisal

• Estimation of willingness to pay (WTP)– Neighborhood specific attributes

• Golf courses, Power lines, Sea view, city plans• Proximity to roads, airports etc etc

– Property specific attributes• Quality, depreciation, size

• Time specific attributes• Index construction

Article 1: House Price Depreciation Rates and Level of Maintenance

1. Introduction– The introduction states the basic objectives of the study and explains why it is

important.2. Literature review

– All papers, even if they are relatively short, should contain a review of relevant literature.

3. Theoretical model and Method– In this section, you describe the general approach to answering the question you

have posed.4. Descriptive statistics

– You should always have a section carefully describes the data used.5. Econometric analysis

– The results section should include your estimates of any models formulated in the models section.

6. Summarize and conclusion– This could be a short section that summarizes what you have learned.

Introduction• Depreciation may bias the estimation of CPI,

appraisals, tax assessment• Why does a property depreciate over time?– Physical deterioration– Functional obsolescence

• Technological changes– External obsolescence

• Changes in neighborhood• My main objective is to estimate different house price

depreciation rates depending on the level of maintenance.– Keeping physical deterioration constant

Literature Review• Malpezzi et al (1987) - literature review, estimate deprecation rates

and how it differs across markets (0.9% to 0.3%).• Shilling et al (1991) - tenure status, lower depreciation rates in owner-

occupied properties• Rubin (1993) - why negative age effect – taste for newer houses• Goodman and Thibodeau (1995) - age induced heteroskedasticity –

depreciation is non-linear, important to include the age effect as a second-order effect

• Knight and Sirmans (1996) – maintenance, similar to my study. (0.9% to 1.9% depending on maintenance)

• Clapp and Giaccotto (1998) - over time, across space• Knight et al (2000) – No effect. My argument is that their assumptions

are unrealistic.• Smith (2004) - across sub-markets (0,5%-7% depending on area)

Method

• Hedonic• Spatial Econometrics (I will come to that)• Specification:

• Implicit price:

Msq AMAAXY 2)ln(

MYAYYA

YMsq

2

Data - Descriptive Analysis Unit Average Standard deviation Price SEK 2545026 1215467 Living area Square meters 118 43 Other area Square meters 58 33 Lot size Square meters 724 266 Rooms* Number 5.03 1.3 Quality Index 27.6 5.9 Sea view* Binary 0.05 Age Year 51 18 Distance Meters from CBD 8754 2694 Sauna* Binary 0.35 Heating* Binary 0.20 Cabel-tv* Binary 0.30 Garage* Binary 0.62 Fireplace* Binary 0.63 Inside maintenance* Binary 0.79 Outside maintenance* Binary 0.50 Drainage Binary 0.38 Electricity Binary 0.56 Kitchen Binary 0.68 Laundry Binary 0.63 Road traffic* Binary 0.28 Number of observations 640

Properties in need of maintenance(%)

Age group Outdoor Indoor Electricity Laundry Drainage Kitchen 0-10 16 37 0 21 0 21 11-20 36 52 20 48 8 72 21-30 39 78 28 64 19 67 31-40 38 79 43 57 36 64 41-50 41 83 53 55 41 64 51-60 63 86 72 70 52 73 61-70 53 84 63 61 39 73 71-80 61 80 67 73 48 71 All 50 79 56 63 38 68

Property age and location from CBD

Distance (meter)

15000-16999

13000-14999

11000-12999

9000-10999

7000-8999

5000-6999

0-4999

Age

(yea

r)80

60

40

20

0

Econometric Analysis

Model 1 Model 2 Model 3 Model 4 Model 5 7 Age -0.01178 -0.01031 -0.00951 -0.00714 -0.00800 (-3.72) (-3.41) (-2.91) (-2.30) (-2.59)

8 Age square 0.00018 0.00016 0.00011 0.00009 0.00011 (5.02) (4.70) (3.04) (2.47) (3.28)

18 IM -0.05429 -0.07818 (-2.08) (-3.15)

19 OM -0.10713 -0.09442 (-5.23) (-4.78)

20 AIM -0.00144 (-3.03)

21 AOM -0.00181 (-5.04) Adjusted R2 0.604 0.669 0.675 0.724 0.725 Moran's I 4.16 2.94 .07 -.21 -0.15

Depreciation effect

-25%

-20%

-15%

-10%

-5%

0%

5%

10%

15%

0 10 20 30 40 50 60

Age effect

IM

OM

IM+OM

10

23

Regression results

Model A Model B Model C Coefficient t-value Coefficient t-value Coefficient t-value Age -0.00807 -2.61 -0.00813 -2.64 -0.00773 -2.51 Age square 0.00012 3.30 0.00011 3.28 0.00012 3.50

AIM -0.00144 -3.03 -0.00046 -1.17 AOM -0.00181 -5.02 -0.00109 -2.42

AElectricity -0.00065 -1.66 0.00023 0.58 AKitchen -0.00139 -3.14 -0.00100 -2.67 ALaundry 0.00017 0.42 -0.00079 -1.56 ADrainage -0.00137 -3.74 -0.00142 -3.84

Adjusted R2 0.725 0.725 0.733

Conclusions

The main contribution in this paper is that the analysis relates depreciation rates to the level of maintenance.

The results show that the depreciation rates are significantly lower for a maintained property compared to a non-maintained property.

The depreciation rate is estimated to be 0.77 percent per year for a well-maintained property and 1.10 percent for a property that is not renovated in- or outdoors year 1.

In year 20 the annually depreciation rate is estimated to be 0.42 percent respectively 0.84 percent.

Article 2: Analyzing location attributes with hedonic model for apartment prices in Donetsk, Ukraine

• Hedonic• Apartments• Values (instead of prices)• Sub-centers

– Monocentric vs. polycentric

• Positive and negative externalities

• Price gradient – different in different directions

Household Consumption Pattern of Housing Attributes

• First stage: House prices as a function of housing attributes hedonic prices

• Second stage: hedonic prices as a function of quantity, income and other socioeconomic characteristics.

• However, problem with the second stage– Simultaneous decision – hedonic price and quantity– Two ways estimating it

• Assume utility function• Multiple markets