1 / 12

# COMPARISON OF RENT PREDICTION MODELS: THE CASE OF ISTANBUL OFFICE M ARKET - PowerPoint PPT Presentation

COMPARISON OF RENT PREDICTION MODELS: THE CASE OF ISTANBUL OFFICE M ARKET. Dilek PEKDEMİR DTZ Pamir & Soyuer. Background. Hedonic office rent prediction models based on multiple regression Difficult to incorporate large number of variables in to a simple mathematical model

I am the owner, or an agent authorized to act on behalf of the owner, of the copyrighted work described.

## PowerPoint Slideshow about ' COMPARISON OF RENT PREDICTION MODELS: THE CASE OF ISTANBUL OFFICE M ARKET' - belle-bradshaw

Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author.While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server.

- - - - - - - - - - - - - - - - - - - - - - - - - - E N D - - - - - - - - - - - - - - - - - - - - - - - - - -
Presentation Transcript

### COMPARISON OF RENT PREDICTION MODELS: THE CASE OF ISTANBUL OFFICE MARKET

Dilek PEKDEMİR

DTZ Pamir & Soyuer

• Hedonic office rent prediction models based on multiple regression

• Difficult to incorporate large number of variables in to a simple mathematical model

• Multicollinearity between independent variables

• Selection of dependent variable; asking or contract rent data

• Aim; to examine the problem with construction of an office rent prediction models and development of a viable prediction models for Istanbul

• Selection of dependent variables

• Asking, gross and net contract rent

• Multicollinearity and reduce number of variables

• Backward method in the standard regression analysis

• Factor analysis to grouped related variables

• Model selection

• R-squared, t-statisctics

• Akaike Information Criteria (AIC) and Shwartz’s Bayesian Criteria (SBC)

• In the light of the literature 34 variables are obtained between 1996 – 2006

• Four office submarkets in the CBD

• Three different rental value

• 59 observations

• 155 contract data is obtained, but only 59 contract data is available for the same office units

• Gross and net contract rent is calculated

• Outlier observation

• High explanatory powers (R2=0.85, adj.R2=0.64)

• Multicollinearity between locational and building variables

• Reduced model with backward (R2=0.77, adj.R2=0.70)

• Gross and net contract rent

• No outliers

• High explanatory powers (R2=0.84, adj.R2=0.63)

• Reduced model with backward (R2=0.79, adj.R2=0.72)

• No multicollinearity in reduced model

• Factor values resulting from factor analysis are substituted into regression model

• 5 factors with eigenvalues explaining 78% of total variance is obtained

• Rent equation is constructed with;

• 5 factors representing the influence of 21 variables,

• 7 independent variables not related to any of predetermined factors and

• 6 dummy variables

• The attributed meanings of factor groups:

• Factor 1; attractiveness for new office investments

• Factor 2; building characteristics

• Factor 3; economic and market conditions

• Factor 4; quality of region

• Factor 5; lease conditions

• Lower explanatory powers (R2=0.48, adj.R2=0.34)

• No multicollinearity

• Gross contract rent

• Lower explanatory powers (R2=0.46, adj.R2=0.21)

• Net contract rent

• Improvement in explanatory powers (R2=0.51, adj.R2=0.29)

• Factor 3 (economic and market condition), office supply and new office investments are found most significant variables.

• No distinctive difference in the explanatory powers (R2) of models with different rental values

• But, the adjusted R2 are improved in the reduced models

• Outlier data in asking rental values while no outliers in gross or net contract rental value

• Multicollinearity between locational and contract variables in standard model, but solved in reduced models

• The explanatory power of models with factor analysis is lower than standard model

• Gross contract rent is more reliable data to produce better rental predictions, it also includes tax effect

• In general, building and locational variables are found significant

• Distance to CBD, transportation nodes, prestigious areas and accessibility are the most important rental determinants

• Secondary centres gain importance

• Quality and prestige of the office buildings; building age, no of elevators, no of floors, parking ratio are found significant

For further information;

Dilek Pekdemir, [email protected]

Hakkı Yeten C., No:12/7, 34365, Şişli/İstanbul, TURKEY

Phone: +90 (212) 231 5530 ext.126