The market of single houses and cottages: Odintsovo district (Moscow Region), cities Novgorod and Tver

(paper was presented at European Real Estate Society Conference-2000 in Bordeaux (France) and published in CD-Proceedings

Yuri Kochetkov, p.h.d.
Center of Real Estate Analysis
E-mail: jorgen@crea.ru

Introduction

    The market of single houses and cottages seems to be a most "obscure" real estate market among others in Russia. In fact, this market is scantily known as compare to markets of apartments or renting premises. On a whole, it is characterised by low activity. Brokers are sceptical of small and average Russian cities single houses markets, which are associated with low liquidity and moderate values. In Russia people used to believe that apartment in the multi-stories building are the best dwelling area, in spite of low prices for the land parcels outside of the city area and the ability of most people to afford cars.
    As compare to apartment, single house is a classical real estate object - land parcel plus improvements (buildings, structures, etc.). For the case of Russia, the following points can be highlighted:


    Besides, initial researches of this market revealed evidence of the four segments (submarkets). First is the market for vacant land, second is the market for old wooden houses with low level of infrastructure, third is the houses which were built in 70-80th (with high quality wooden construction or brick/block walls). Forth is the market of contemporary large houses (gross building area greater than 100 sq.m., at least, two stories) with good infrastructure and high quality finishing. People used to call these houses "cottage".
    The single houses markets can be considered from another point of view - by regions. No doubt, specific features of regions dominate over other market factors. The most developed market is the market of Moscow and St.Petersburg regions. In this case, the offers for houses, cottages, and parcels outside the city range dominate over ones inside the city. On the contrary, average and small Russian cities have another market. This market operates by parcels and houses inside the city range. Vacant land is rare case for such market.
    The main tasks of research are as follows: confirm the hypothesis of existence of several submarkets and estimates the difference of single house markets for two regions - central and provincial. This research is the pioneer work and, mostly, quantitative characteristics will be considered in this paper.

1. Model building of market for single houses and cottages of Odintsovo district (Moscow Region)

    Odintsovo district is unique for Russian real property market due to this district is the most prestigious area near Moscow for summer time rest and recreation and for residential outside of the capital. Good ecology, vast recreation resources, well-developed communications define the highest land price here. For instance, land price near Barvikha and Razdory was about $200/sq.m. at the pike of real estate value increase. It should be noted, that this high price is the result of such factor as close the distance to mighty ones residence (president, ministers etc.). It's very problematic to take into account this factor and it's very hard to get the correct information about such sales.
    Database consisting of 250 records has been collected, but some records were excluded due to lack of factors information, incorrect price etc. Primary information about market (brokers opinions etc.) has been collected in 1998 before data collection.

       1.1 DATABASE

    Database was divided into two parts:

A. Sales for land parcels and single houses within the range 1995-1998 years

B. Listing prices for land parcels and cottages at summer 1998 year

    Information about collected factors and parameters for these two sample is summarised in Tab.1

Factors & Parameters data availability ,

statistics in sq.m.

Sample A

Sample B

Av.

min

mean

max

Av.

min

mean

Max

Parcel square

+

250

1460

6000

+

600

1248

2400

Total area of building

+

22

131

524

+

175

327

1400

Living area of building

+

18

73

236

-

      

Square of underground floor

+

      

-

      

Communications lines

+

      

+

      

Access to roads

+

      

+

      

Wall type

+

      

+

      

Finishing of buildings

+

      

+

      

Number of floors

+

      

+

      

Additional constructions

+

      

+

      

Settling of parcel

+

      

-

      

Veranda/Attic/Balcony

+

      

-

      

Garage

+

      

+

      

Telephone

-

      

+

      

“Ready for living”

-

      

+

      

Coordinates

+

      

+

      

Date of sale (offer)

+

      

+

      

Record numbers in sample

144

      

52

      

    It was initially clear, that models for samples A and B should be built separately due to differences in the collected factors and parameters. Thus, sample A consists of ordinary houses with considerable deterioration, median building area - 91.6 sq.m., 1-2 stories, houses are evenly distributed in the district. The sample B, in turn, consists of contemporary houses (so-called "cottages"), median building area - 360 sq.m., 2-4 stories, houses are concentrated in the eastern part of district (near Moscow).
    Such situation allows us to perform comparisons between markets of single houses and cottages, and to find the differences between model building process for offers and sales. Initial hypothesis was "there exist a great differentiating among these markets" and this one based on the following ideas:



       1.2 ANALYSIS TECHNIQUE

    Standard mass appraisal technique (CAMA), developed by American specialists in 70-80th, has been used (Eckert, 1982). In this technique main trends and tendencies can be determined by means of market model building. During the process of model building, the general model parameters: influence of factors, non-linear trends etc. are calculated on the base of Multiple Regression procedure, Descriptive Statistics, Cluster and Factor Analysis, and Feedback. Normally, three groups of factors are considered: 1) the property characteristics, 2) the location factors, 3) the time trends. Final coefficients can be figured out by Feedback or Regression procedures.
    In our case - during market analysis of Odintsovo district, considerable difficulty has been experienced in regression analysis due to low ratio of the number of records to the number of factors (2.36 for sample A and 2.26 for sample B). That is why the Feedback became the main technique for model building. Coefficients calibrations has been performed in the frame of the following hybrid model structure:

Paper: Real Estate Market of Moscow region, Russia

where RS - location adjustment; T - time trend adjustment; kgen - overall adjustments on the property characteristics; kbld - building adjustments, Stot, …, Sadd - squares of building, parcel, and additional structures; kpcl - land adjustments; kadd - additional structures adjustments; B1-j - base rates.

    Response surface (location additional structures) has been built by ratios technique 1, time trend has been figured out by this technique too. Statistical system NCSS6.0 was used for this purposes.
   The results of model building are the response surfaces, time trend, base rates.

       1.3 RESULTS AND DISCUSSIONS

    The main difficulty feature of market analysis for sample A is a difficulty of correct calculation of ending weights (adjustments) in the hybrid model. This is the result of imperfection of data, lack of important information (for instance, factor of prestigious neighbourhood), and, sometimes, incorrect prices. Nonetheless, plausible, logical model A has been built under circumstances that less than 10% of records were discarded. The main parameters and results for this model are shown in Tab.2-4 and pic.1-2.

Parameter Value
Land parcel $4820 /100sq.m.
Total area 2 $990 /sq.m.
Hybr. total area 2 $840 /sq.m.
Hybr. living area 2 $235 /sq.m.
Underground area $24 /sq.m.
Tab.2 The values of base rates for main area units for model A

Factor Adjustment
Brick wall 1,64
Wooden wall 1
One-story house 1,06
Two-story house 1
Undeveloped parcel 3 1,61
Tab.3 The adjustments values for some interesting factors for model A

Stat.coef. Value
R2 0,86
COD 39,6 %
Tab.4 The values of some statistical coefficients for model A

Paper: Real Estate Market of Moscow region, Russia
Pic.1 Time trend for model A

    As for sample B (listed prices) - there were no any problems with the model building process due to absence of sale conditions factor and initial correctness of listing information.

Parameter Value
Land parcel $3050 /100sq.m.
Total area $1810 /sq.m.
Tab.5 The values of base rates for main area units for model B

Factor Adjustment
"Ready for living" 1,83
Two-story house 1,19
Three-story house 1,01
Four-story house 0,78
Undeveloped parcel 3 1,27
Tab.6 The adjustments values for the number of most interesting factors for model B

Stat.coef. Value
R2 0,87
COD 17,4 %
Tab.7 The values of some statistical coefficients for model B

    Model differences for samples A and B as well as good quality of predictions for sample B (see statistical coefficients R2 and COD) are not a surprise. For instance, listed prices are based on the limited number of factors (normally, sets of these factors are published in listing) and the values of these factors is more unified. These features make listing database attractive for model building. However, we should remember that here point of view of seller (or broker) is dominating over point of view of consumer. Moreover, there is some listed prices increases (5-15%) in expectation of possible bargaining with clients. No doubt, databases of sale data are more preferable from the point of view of price prediction, but listing databases can be used in general market analysis and calculation of factors contribution in the property value.
    Let us concentrate on the difference of base rates for GBA and parcel. Predicted land value in the model A is 50 percent higher than in model B. Phenomenon is unexpected. Normally, listed prices slightly more than sale prices. But in our case we should remember our hypothesis about existence of two submarket. In Russia, size of land parcel is not important for prestigious cottage owner. Moreover, a large share of cottages are located outside of villages and hamlets - in the so-called cottage communities. As for ordinary householder (mostly living in villages and hamlets), size of parcel is very important and buyers are ready to pay for each additional square meter. These circumstances are determining factors about the differences.
    On the contrary, predicted value of GBA (general building area) sq.m. for cottages twice as large as value for ordinary house. This fact is expected - replacement cost for cottages is considerably higher than for ordinary house. Besides that, prestigious factor is reflected in listing prices. Note, value of GBA sq.m. for unfinished cottage is about 45% below the GBA sq.m. value for "Ready for living", so this value is closer to the GBA sq.m. value for ordinary houses.
    Let us discuss one interesting phenomenon - factor of vacant land. In Russia it was firstly discovered during model building for Novgorod single houses market in 1998. Phenomenon is the following: value of sq.m. for vacant land is more than value for land at the house. At this time two explanations are put forward for this result:



    Some adjustment differences on the vacant land factor for models and B can be explained by difference of acquiring rights in the case of village (A) and cottage community (B).
    Interesting result has been obtained for adjustment on the number of stories for model B. Analysts used to believe that the more stories the more prestige for cottage owner under circumstances of the same GBA, but it seems to be not right. Perhaps, in Russia tendency to have house of 2-3 stories for one family has get prevalence. For model A this adjustment is insignificant, but has the same trend. Additional investigations are needed to know - what is the nature of this phenomenon.
    It remains to consider two another important factors - location factor (response surface) and adjustment on sale date. Time trend has been obtained for sample A only because of collected listed prices (sample B) relate to narrow time period of the Summer 1998 (before well-known crisis). Time trend doesn't display season fluctuations - either fluctuations were not formed or sample A wasn't representative. However, two tendencies for Russian single house market should be noted on the whole:


Paper: Real Estate Market of Moscow region, Russia
Pic.2 Response surface for model A, (marked: MKAD - Moscow Circle Highway; Od. - city Odintsovo)

Paper: Real Estate Market of Moscow region, Russia
Pic.3 Response surface for model B, (the same scale as for Pic.2)

    Response surfaces are shown on pic.2-3 for considered submarkets. First of all, a difference in the scale of location adjustment should be noted: for model A the ratio between maximum and minimum value of this adjustment is 8.5, for model B - 15.5. It's very simple to calculate that in the best location the land price is equal - Principle of competition is working. The difference of scales is determined by factor of prestige, rich consumer is looking for cottage in the best place, but ordinary one is ready to live anywhere. Poor places are not interesting for cottage buyers (well-known principle: the good place - the expensive house).
    The difference of submarkets makes itself evident in the formation of the centres of positive attractions (CPA) . In the Odintsovo district these are some communities. On the both surfaces the main CPA (No.1) is north-east area of the district (where President Yeltsin and some other mighty ones are living). As for second one (No.2) - this CPA is different for submarkets, for model B this is area near v.Nikolina Gora (Head of Moscow Luzhkov living complex is near this village). For model A CPA-2 consists of two close local CPAs in the south direction from Nikolina Gora. Village Nikolina Gora itself is quoted slightly greater-than-average among permanent residents. Unfortunately, shortage of data doesn't permit more detail investigation of the location influence on prices. Response surface techniques did manifest the main attractive centres and pointed on the additional differences for submarkets.
    On the whole, we can note that the market of Odintsovo district is logical, rational and can be numerically analysed. This market includes, at least, two submarkets - a) ordinary single houses in the villages and hamlets; b) cottages in the cottage communities. These submarkets are competing one with another in some locations.


2. Market of single houses and cottages for cities Novgorod and Tver

    Tver and Novgorod are the typical intermediate cities of the central part of Russia with medium-developed single house market. Russian consulting company CREA (http://crea.ru/engcrea/) performed the analysis and model building of local markets within the frame of Tax Reform Project (USAID) in 1996-1999. Developed models were applied to fiscal cadastre database during assessment process. The results of this work were published in the number of project reports, here we will show the general results only (see also (Kalinina, 1998)).
    The mention above hybrid model structure was used for model building for both cities. Coefficient calibration was done by means of feedback procedure. The models quality was medium, COD was 17-21%, but models were upgraded several times, so latest field reviews demonstrated good work of the models. Below (Tab.8-12) the main base rates and adjustments are shown. Vacant land sales records in Tver and cottages sales records in Novgorod were not included in samples.

Parameter Value
Land parcel $9.50 / sq.m.
Cottage total area $410 /sq.m.
Single house total area $295 /sq.m.
Tab.8 The values of main area units for Tver single houses and cottages model (Summer 1998)

Factor Adjustment
Brick wall 1,41
Wooden wall 1
Tab.9 Adjustment for kind of wall for Tver single houses and cottages model (Summer 1998)

Parameter Value
Land parcel $8.90 /sq.m.
Single house total area $225 /sq.m.
Tab.10 The values of main area units for Novgorod single houses model (Spring 1998)

Factor Adjustment
Brick wall 2,5
Wooden wall 1
Undeveloped parcel 1,57
Tab.11 The values of adjustments for some interesting factors of Novgorod single houses model (Spring 1998)

Tver model Novgorod model
3,3 4,0
Tab.12 The range of RS values (or ratios) in the best location of city to the worst one for corresponded models

    First of all, let us consider land value. As may be seen from Tab.13 the land value of Odintsovo district considerably dominates over land value of mediate Russian cities. Undoubtedly, this is the result of neighbourhood of the Russia capital and influence of such factor as prestige. People are ready to pay greatly more to live in Odintsovo district (it should be mentioned that Tver region adjoins to Moscow region). It's very interesting that RS value ratio between the best location and worst one in Odintsovo district is greater than this ratio in Novgorod and Tver. It is the influence of prestige factor again.
    As for buildings, Odintsovo district prices are considerably more than ones within the cities. It is interesting that for Tver ratio between ordinary house value and cottage one is low than in Odintsovo district. However, on a whole, the adjustment values on factors either close one another or works uniformly for all market models. This fact says that the universal markets laws are working for all regions. Quantitative difference for some adjustments (for instance, on the kind of wall and vacant land factor) can be result of local peculiar features.

Land parcels Single houses Cottages
~ 4,3 ~ 3,8 ~ 5
Tab.13 The ratios of the main area units values of Odintsovo district to means of main area units values for cities Tver and Novgorod


Conclusion

    Based on initial analysis, hypothesis of coexistence of two single house submarkets is confirmed at least in Odintsovo district. The market of ordinary houses differs from the market of cottages by several quantitative parameters, such as base rates for land and building areas, mean values of building characteristics and response of the price to location. Besides that, a considerable price gap between Odintsovo district and Russian center cities has been observed. Another important result is that market of single houses in Russia can be subject of model building and quantitative analysis.


Supplement
    An actualisation of results to 1999 year.

    The described above market is related to before-crisis period. The Crisis of August 1998 has affected all Russian real estate market as a whole, in some regions prices in hard currency have fallen to half at the end of year (common feature for large and average Russian cities - prices is calculated in dollars, not in roubles). Drop in prices touched Odintsovo district too. One-two autumn months listing prices were not changed, but absence of transactions made drop in prices inevitable. This market became stable at autumn of 1999 only. So, it is possible to perform actualisation of results with the help of ratio approach. To solve this task some additional records at September 1999 have been collected and ratios of prediction to listing price have been figured out. Results are shown in Tab.14.

Property Prediction by
model Sep.1998, $		 Listing price
Sep.1999, $		 Ratio
  A C C/A
Cottage in v.Lapino 129433 70000 0.54
Cottage near Odintsovo 631771 205000 0.32
Cottage in v.Zhavoronky 95437 65000 0.68
Cottage near Barvikha 288746 140000 0.48
Cottage in v.Zhavoronky 220631 190000 0.86
Cottage in v.Shulgino 28400 24000 0.85
Cottage in v.Mamonovo 109860 75000 0.68
Tab.14 Predictions, listing prices, and ratios for Autumn 1999

    The mean value is 0.57, median value - 0.68, so we can conclude that since autumn 1998 prices dropped more than a third of its value before crisis. Thus, to apply results of model building in 1999 we should reduce all base rates by 35-40%. However, this estimation is not reliable due to small number of objects in the control sample. More detail and data collection (desirably - sales) should be done to reveal some current tendencies in Odintsovo district market.


References
  e.g.1 J.K.Eckert (gen.ed.), 1982, "Property Appraisal and Assessment Administration", International Association of Assessing Officers, USA

e.g.2 N.Kalinina and Yu.Kochetkov , 1998, "Computer Assisted Mass Appraisal in Russia: first results", Scientific Park, No.1 (in Russian)
Footnotes.

1This technique based on the calculation of a ratio of sale price to predicted price for all records within the frame of model without factor which is considered. The deviation of the ratio from 1 is assigned to the influence of this factor. Next, dependence of ratio on factor value is approximated by non-linear function.

2The base rates for hybrid areas are calculated within the frame of standard hybrid model (all property areas are included). The base rate for general building area is calculated for the simplified model, where this kind of area is included only).

3Factor of vacant land in listing.

Counted by L-STAT