The impact of new metro stations on rental prices in Moscow

Размещено на http://www.allbest.ru/

The impact of new metro stations on rental prices in Moscow

Table of contents

Introduction

1. Literature review

2. Rental and real estate market in Moscow

3. Predictions on the effect on price

4. Methodology, data and empirical results

Conclusion

List of References

Аннотация

Московский метрополитен расширяется с каждым годом, оказывая влияние на многие рынки, в том числе, на рынок аренды и недвижимости. Улучшение транспортной доступности отражается на ценах на жилье. В данной статье анализируется литература по теме влияния транспортных структур на стоимость недвижимости и арендную плату, описывающая все современные методы исследования. Она демонстрирует тенденции Москвы 2018 года, а именно рост цен на жилье и снижение арендной платы. Прогнозы влияния на арендную плату были сделаны через максимизацию полезности для арендаторов. Они констатируют обратную зависимость расстояния до ближайшей станции метро от цен, но оставляют общий эффект от строительства новой станции метро неоднозначным из-за наличия косвенных затрат. Эмпирически эта статья подтверждает, что в Москве существует разница между предложениями аренды квартир, которые уже имеют доступ к станции метро, и теми, кто получит его в 2018/2019 годах, демонстрируя, что уже действующая станция метро поблизости повышает арендную плату в среднем на 8,1%. Было установлено, что природные объекты, такие как парки и озера, не имеют большого влияния на цены на аренду, в то время как негативные эффекты, такие как загрязнение окружающей среды или шум, вызванные новыми станциями метро, не улавливаются близостью к ним. Кроме того, эмпирическое исследование подтверждает обратную зависимость расстояния до ближайшей станции метро и арендной платы. Однако многие арендаторы были осведомлены о будущих выгодах и использовали эту информацию в публикуемых объявлениях, таким образом, цены могли вырасти более чем на восемь процентов в общей сложности.

Abstract

metro value rent transportation

Moscow metro is expanding each year, having profound effects on the rental and real estate market. Improved transport accessibility is reflected in housing prices. This paper analyzes literature on the topic of influence of transportation structures on property values and rents describing all modern methods of research. It uncovers trends for Moscow of 2018 that establish increasing housing prices and diminishing rentals. Predictions for the influence on rents are made through tenants' maximization problem. They state inverse relationship of distance to the closest metro station on prices, but leave the total effect of construction of a new metro station ambiguous due to the presence of indirect costs. Empirically, this paper confirms that in Moscow there is a difference between the rental offers for apartments that already have an access to a metro station and those who will get it in 2018/2019 demonstrating that already functioning metro station nearby increases rental prices on average by 8.1%. It was found that natural amenities do not bring much value to rental prices, while the negative effects like pollution or noise caused by new metro stations are not captured by proximity to them. Moreover, empirical study confirms an inverse relationship of distance to the closest metro station and rents.

However, many renters were aware of future benefits and exploited this information in offers, thus, prices could rise more than by eight percent in total.

Introduction

In recent years, the capital city of Russia has been actively developing its transport infrastructure, primarily the metro. This fact is not surprising as passenger traffic on the Moscow metro is one of the highest in the world. In terms of the number of passengers carried per year, it concedes only to Beijing and several other Chinese subways. For example, in 2018 the average daily passenger traffic was 6.668 million people. To ensure good mobility in the city, many new stations are being built. According to the capital's construction complex, in 2018 the total length of metro lines increased by 33 km, 17 new stations were opened and three depots were built. In February 2018, Moscow mayor Sergei Sobyanin said that 50 more new stations are planned to be built over the next five years. In 2019 8 new stations were opened, in 2020 - 6 metro stations. These new stations may change the way different markets function including real estate market. Moreover, they can have profound effects on commerce and retail in the affected areas.

For those who are planning to buy an apartment or just rent a flat in Moscow, the presence of a metro station nearby can be the most important factor in the attractiveness of housing. The appearance of a new station can increase the cost of housing because transport accessibility increases significantly as a result. Metro construction affects real estate prices and rents, but also can lead to the construction of new buildings -next to the stations that will open in the next two- three years. New retail possibilities will appear, and other amenities like restaurants and coffee shops or even shopping centers will open getting an opportunity to grow for business. All of this increases the attractiveness of an area and these factors feed off each other.

Improved transport accessibility is reflected in housing prices. According to real estate market specialists, the influence now diminished in comparison with the situation ten years ago. If in 2010 the appearance of metro in Butovo ot Mitino could increase housing prices by 20% (Julia Ryshkina, IRN.RU, 2019), nowadays this effect is “flatter” and is estimated at around ten percent or even lower (according to some reports, around five percent). Any additional factor that increases the already high demand could give a very noticeable rise in price. At the same time, prices do not rise immediately, but gradually - first when clear plans are outlined that a new station will appear in the next year or two, then a little in the course of construction and another stage of price growth is noted shortly before the station opens. That is, even such a small growth is "smeared" for two years (conjecture is based on the analytics by real estate agency Azbuka Zhilya). In addition, price increases occur primarily where the appearance of new stations really greatly improves transport accessibility. It is true for the areas where there was no metro in principle (for example, in Nekrasovka or Kommunarka), so these areas can help to catch the highest effect from new transport opportunity.

The impact of metro construction on the dynamics of prices for new buildings is quite difficult to track, because developers initially put this factor in the cost of housing. However, of course, the appearance of the metro drives up the demand for real estate, which allows developers to increase prices more actively as the stage of construction is approaching.

The aim of this paper is to primarily measure the impact of a construction of new metro stations on rents as it will exclude opportunities of developers and will focus on capturing the trends created by market presented by renters who will react to the news provided by mayor. In other words, the main idea is to measure the effect of transportation infrastructure and land-use characteristics and to investigate the impact of the announcement of a future metro stations on the area. Research will be based on rental offers published in 2018 that account for announcements on 2018 and 2019 openings of stations.

The work is organized in the following way. Firstly, the literature on the topic is analyzed to study the experience of different cities from around the world. This chapter carefully describes the methods of research on transportation systems and their influence on housing prices. Moreover, it helps to make predictions on the empirical result that can be figured out further. Secondly, trends of the property and rental prices in Moscow are described to explore the mood of the market. Thirdly, using the modified Alonso model that can be useful in terms of predicting impact of proximity to some factor, the effect of distance to the closest metro station and other factors is investigated through the maximization problem of consumers (those who are looking for a housing unit). Finally, the data on Moscow rents is presented and empirical research is done with the Eviews statistical package for an econometric analysis.

1. Literature review

The idea of investigating the effect of such type is relatively new, it starts to develop only in this century and it becomes more and more popular nowadays that is why most of the articles that will help us to analyze the case of Moscow. As the research we can operate appeared recently, the methods of the investigations are modern and they don't need much adjustment due to the progress that is made over time. However, the primary target is to catch the relationship between housing sector and transport systems that function in different cities and countries. It is important to understand that the location of metro lines, the number of residents, the quality of construction of houses and many quite unexpected factors completely change the picture and can give different results in an attempt to predict changes in the real estate sector.

The first modern method that should be mentioned is difference-in-difference models that account for main and interaction effects. It was done in the sphere of real estate development from announcements before the opening of light rail transit system (Cao, Porter-Nelson, 2016). It exploits a treatment-control approach: the corridor where the line operates is taken as a treatment, but for the control four additional corridors are used, high-frequency buses run in these areas. Authors consider difference-in-difference models much more powerful than crosssectional models in the context of testing announcements on the panel data that include monthly counts and values of building permits in the corridor of thirteen stations as a treatment and fifteen bus stops as a control. The main results confirm no impact of announcements that refer to engineering and positive impact of announcements that refer to funding on permit values by 80% and on number of permits by 24%. These findings highlight the importance of respecting the timeline in our research on Moscow.

A multilevel model can be considered a good model to investigate the changes in residential housing prices (Mulley, Tsai, 2016). Comparing the price change before and after the introduction of a new transport infrastructure is not enough - catchment and control areas may become incomparable over time due to the change in land usage or due to some other factors. Using a quasi-experimental approach is better but it also does not take into account property attributes and neighborhood characteristics. Multilevel modelling solves these problems and it allows hierarchical or clustered structure. While controlling price determinants and time effects, this article proves a property value uplift caused by Bus Rapid Transit system. The important point is the competent work on statistical errors that can be underestimated in OLS in case of hierarchical data which is done by this type of model. This research also helps to understand how the significance area may be revealed.

One more method is a spatial quantile hedonic model which is tried in the analysis of the impact of Shanghai metro on rents. It is compared to conventional back-to-the-mean hedonic pricing method that gave less robust and comprehensive transit premium estimates. The resulting finding is that there is no evidence of transit-induced segmentation of the local rental market (although observing a significant rent premium for transit proximity, variation is the clue factor), but the positive effect can be seen for two-bedroom-one-bathroom apartments' rents. The research brings the idea of making a good economic intuition through the overview of residential property market which is booming in this case and where the demand always exceeds supply. It also provides a list of variables that can be considered in the studies of such type including case of Moscow - community age, greenness, hospital, metro station, park, school, urban center and weighted rent. One more important conclusion is to use rents as a more stable dependent variables than sale prices, because property prices include “speculations on future capital gains which cannot be explained only by improved accessibility due to investment in public transportation facilities” (Wang, Feng, Deng, Cheng, 2016).

The case of Wuhan as one of the most recent studies (Tan, He, Zhou, Xie, 2019) may be interesting to us to different extends. First of all, it takes already familiar method of difference-to difference model and combines it with traditional hedonic price model to control for the effects of accessibility, housing attributes and landscape amenity. The data was collected in the form of remote-sensing images by a change-detection method. The work highlights the existence of both positive and negative effects where positive ones are quite apparent while negative ones are presented by created noise, environmental pollution or increasing crime rates. It underlines the direction of deviation for price premium towards overestimation in case of unobserved factors related to infrastructure including joint effects. The most valuable information concerns the importance of taking into account educational component, which is crucial for parents and significantly boosts the price, and impacts of natural amenities like lakes, rivers and parks that have the same influence on rents. It is essential to differentiate between transfer and non-transfer stations as this research shows higher demand on property near transfer stations and the extended effect of the distance from the station that adds value. The proximity to city center is best caught by difference-to-difference model that is perfect for the situation of the presence of trend (increasing price) even beyond infrastructure changes. Overall, the models confirm positive externality created by closeness to metro station, they control for housing characteristics, accessibility, natural environment, time and district fixed effects and prove the decreasing impact of new stations with the increase in distance. However, it can be said that effect was heterogeneous as greenfield stations had a clear impact on land revitalization, while infill stations were not so “influential”.

“Access to goods and services is the main factor influencing the development value of land” (Alonso, 1964), however, some studies show negative correlation of distance to stations and property values (Pan, 2013), that is why there was conducted a research on the Sydney Northwest Metro Line. This work (Chen, Yazdani, Mojtahedi, Newton, 2019) pays attention to different stages of transportation projects which are pre-planning, planning, construction, and operation and shows that impact on property prices is negative at the announcement but positive at construction stage. It mentions that generally the impact varies also depending on property type as it is positive for the multi-family property and the opposite for the one that is singlefamily. The study has a clear classification of methods that are used - the hedonic price model. The main idea is that property value is determined by different attributes such as location attributes, structural attributes and neighborhood attributes, all of them are capitalized into prices. The strongest influence comes from structural attributes such as the presence of parking, number of bathrooms etc. This research uses three forms of the model - linear form, log-linear (elastic) and semi-log linear (growth) form, where the elastic model is proved to be the best performing one. Result conforms previous studies meaning the negative correlation with distance and property price uplift from the construction of new stations.

Case of the new Line 4 of the Santiago metro system (Agostini, Palmucci, 2008) exploits Alonso model (1964) to show that property prices should be negatively related to the distance from the nearest metro station, this exercise seems to be very useful as it covers the utility maximization of consumers subject to transportation costs and their effects on housing prices. To verify this prediction, authors use hedonic price estimation and difference-to-difference estimation, using a large group of explanatory variables that include structural characteristics of apartments, access to public and semi-public goods and dummies that capture fixed effects for months, years and community. Moreover, to distinguish between treatment and control special variable for a 1000- metre range from metro station is introduced. In terms of the absence of some data for new motorways as a competitor, the check for robustness is done excluding the property close to competitors. Health and education organizations surprisingly are insignificant, one of the possible explanations is that quality of service can be more important than any distance. In brief, a new metro line is a semi-public good that reduces the cost of travelling to work and shopping that is why the obvious result of capitalization of benefits into prices is obtained even with the main focus on the announcement of construction. These results are used to estimate the increase in tax collection.

The study from Greece (Efthymiou, Antoniou, 2013) demonstrates one of the most interesting cases as the result is not so predictable. First of all, this work criticizes hedonic price model for the assumption that “dwellings belong in the same real estate market”. Instead of that, it introduces spatial econometric models that take into account spatial heterogeneity, spatial dependence and include lags of variables. The research itself is conducted using spatial autoregressive model, spatial autocorrelation model, spatial error model and a linear regression that is transformed using a Box-Cox method in order to “stabilize the variance and result in better correlation between the response and explanatory variables”. According to their outputs, spatial models outperformed linear regression as they removed spatial autocorrelation getting the result that was closer to the real one. Authors note that property prices initially are higher for the houses constructed before 1950 due to the architectural interest to neoclassical period. The reason why this particular case can be interesting is that property in the zone of 500 m near the stations had lower price while rents were unaffected at all. Paper suggests possible explanations of these phenomena: for the purchase price, they are lower due to the negative externalities created by construction of a new line keeping demand at the low level because of expectations that consider construction to not be completed within next five years; as for the rents, they are assumed to be short-term from tenants' point of view so the future benefits do not bring value to them. Rationale is based on the survey within owners of shops who claim that they are in difficult economic situation because of construction works and on insignificance of rents' coefficients. However, with the same construction company the results for already operating lines is different - it demonstrates habitual increase in prices and rents. Thus, people treat periods of construction and operation in different ways.

Case study from Beijing draws attention to local specifics meaning that public transport for Asia is more crucial than for Western countries so passengers may be more sensitive to the changes. This research (Li, Chen, Zhao, 2017) criticizes traditional hedonic models for failing to account for spatial autocorrelation and omitted variable and for impossibility of seeing the impact of one variable on housing prices in different locations. As the previous research from Greece, they prefer spatial models that help to overcome these problems. It is important to note that this critique does not necessarily mean that hedonic price models always fail, some studies show the identity of results (Martinez, Viegas, 2009). Paper on Beijing is considering spatial error model the most efficient due to full representation of spatial dependence and capturing shocks on unspecified variables. It investigates metro headways, access to different metro lines and accessibility to job opportunities via public transit within no more than one hour. All the metro service variables including employment accessibility were found to be significant setting the value uplift for houses. People are ready to pay more, and the given intuition suggests improvement in job search scope and greater premiums near workplaces. Moreover, it suggests shorter headways for newly built stations, and thus, shorter waiting time which can be capitalized. Metro station proximity was also found to be valuable. This work mentions some limitation that it is impossible to judge for the causal relationship basing on cross-sectional data that do not account for the changes before and after metro development.

Modified repeat sales model can be also used to verify the significance of intertemporal changes of rent gradient. By comparing time trends for different groups, it is possible to see the positive marginal effect for at least 100 meters distances from new metro stations (Lee, Ryu, Choi, Kim, 2018). In fact, the research from Seoul was concentrated on both accessibility effects - on overall network connectivity improvement and on local accessibility. The data was analyzed through ArcGIS technologies and through the model that used average network distance that was calculated on the base of the first year (2000). In turn, repeat sales model exploits an inverse-log time dummy hedonic model. Results show that “even without an increase in ridership, existing subway users can enjoy a better service level and be willing to pay more rent for a closer location to a subway station”.

To sum up, modern methods include difference-in-difference models, multilevel models, spatial quantile hedonic price models, traditional hedonic price models, repeat sales models and spatial econometric models such as spatial error model etc. Of course, cross-sectional data has more limitations as it does not allow to consider the changes before/after metro development.

Variables that can be exploited are community age, greenness, hospital, metro station, park, school, urban center and weighted rent. One more important conclusion is to make a research on rents and not on sale prices, because property prices include “speculations on future capital gains which cannot be explained only by improved accessibility due to investment in public transportation facilities”. Overall, when catching the effect of transport development on rental prices, the worldwide results confirm the capitalization of new benefits and prove the existence of positive externality. The most interesting problem for our research would be verifying the causes for Thessaloniki and its negative result which is, as we see, very unusual for such type of investigations. However, it can be said in advance that with high probability our result will show positive impact of construction of new metro stations in Moscow on rents looking at the observations from all over the world.

2. Rental and real estate market in Moscow

Since its inception in the first half of the 1990s, Moscow's modern real estate market has been consistently highly attractive to investors and private buyers. According to various studies, the capital is one of the most expensive cities in the world in terms of real estate prices. This applies both to the apartment market and to the commercial and suburban real estate segments.

According to the 2020 census, the population of the Russian capital is 12,692,466 and continues to increase, primarily due to migration. This circumstance, as well as the status of the capital region as a strategic center of social and economic development, implies a high return on investment in real estate in Moscow and accordingly, high demand that pushes up prices resulting in the growth that sometimes slows down and sometimes accelerates:

Fig. 1 Russia 's house prices growth

Source: CEIC Data

Due to increased competition among developers, the level of housing provided is also increasing. More and more often, for example, there is a panoramic glazing of houses, various innovative solutions are installed in apartments, courtyards without cars are being improved. The service of comfortable housing gradually becomes available in Russia. The terms of purchase are improved and favorable offers are provided for installments. At the same time types of houses changed dramatically. The first monolithic houses appeared in the 90s and were initially positioned as elite, respectively, and the cost of a meter in them is still quite high. This is due to the use of new technologies and high construction speed. Nevertheless, prices rise relatively evenly for all types of housing (the directions of change coincide almost always):

Source: IRN.RU

Property prices include “speculations on future capital gains which cannot be explained only by improved accessibility due to investment in public transportation facilities”, that is why it will be crucial to concentrate more on rents. The real estate rental market in Moscow and the Moscow region is also changing constantly and at a rapid pace. Apartments of minimal size are in demand in all classes of residential real estate. One-room apartments in a new building on the Moscow market start from 16 square meters and demand for this segment is growing, especially among renters, due to their liquidity. This moment will especially affect the residential real estate rental market. The timeline for data on which the research is constructed is assuming the year of 2018.

Trends of that time show that housing rentals in Moscow were falling in price. Offers depending on flat areas fell by 1-4% compared to the previous month. Moreover, the cost of rental real estate has decreased in proportion to this parameter and to the class (low segment, comfort, high segment). The exception was one-room flats low segment. Their prices rose by about 0.5%.

These data were announced by analysts of the residential real estate rental market. “For example, a 1-room economy class apartment in Moscow could be rented on average for 25 thousand rubles a month. 2-room apartments, having demonstrated stagnation in June, in July became more affordable by 0.9% - 28 thousand rubles while 3-room apartments lost 2.3% in value to 38 thousand rubles” (Mamontova, 2018). However, in 2018 a one room economy class apartment was affordable for 20 thousand rubles whereas a two-room apartment - for 24 thousand rubles or even lower. A similar trend was registered in the comfort class where one room apartments rose in price while flats with more rooms dropped. But in the high segment, prices for rental properties were adjusted only downwards (they became cheaper by 1-3%).

The crisis in the economy has affected the rental market in Moscow. Comfort-class tenants, who now found it more difficult to pay for such an apartment, were moving to the economy segment. Therefore, the demand for the cheapest objects fell the least, and therefore, prices for them were staying constant or increased. One more trend concerns the number of rooms - those who lived in three-room apartments, were moving to two-rooms; those who lived in two-room apartments - to one-room flat, whereas those who were living in one room apartment, started to rent rooms and share rentals. The latter became a widespread practice and it is still present nowadays.

Regarding rented high-budget real estate, experts argued that prices have decreased due to the stabilization of the ruble to the dollar. Despite the fact that many owners value expensive apartments in dollars, they charge their tenants in rubles. As for the predictions for future, in 2018 analysts and experts predicted a further strengthening of the trend of cheaper rental of quality housing in Moscow.

Thus, at the time for which the research is conducted, property prices had an increasing trend and wise versa, rental market demonstrated diminishing prices that were predicted to fall further. At the same time renters were moving to apartments of lower class and with smaller number of rooms.

3. Predictions on the effect on price

Bid rent theory predicts that price for real estate changes with respect to the distance to the city center. The main idea is that people compete for the land, it is based on the utility maximization and states the willingness to pay more for being closer to center. This type of land-use structure was investigated in the Alonso model from “Location and land use” (1964) where behavior of firms and households is explained. In this model the city is seen as a central business district around which workers settle, and there is competition for land between various uses: offices, shops, and housing. Firms and households have their own rental rate function-the willingness to pay for a location relative to the city center. The model focuses on profits for the firms and on distribution of household income. It allows to study the location and population density in cities. Results demonstrate that the growth of average incomes of the population leads to a concentration of the poor in the central part of the city, as in connection with infrastructure development, travel comfort and increased utility of living in large areas, the discomfort tends to zero, which means that the rental rate curve for households becomes flat. Prices in the suburbs rise due to the preferences of rich residents, while prices in the center are falling. Transport accessibility begins to act as an inferior good, as the demand for it decreases with increasing income. This model is used to show that the process of suburbanization is developing, which primarily determines the decrease in population density, rather than population growth.

Alonso model can be transformed in order to show the effect of distance of rents by using the same technique of simple profit maximization. For this case we should examine only households and their decision of location and size of a house unit regarding metro stations. Even when the model is quite old, it is still applicable when it uses simple mathematical rules like for a simple capitalization model.

Let's introduce the model. It should contain utility, property size, distance as in classical model but at the same time we can try to include indirect costs that are negative (noise, pollution) and positive (access to any possible activities available by metro, reduced time of waiting). In turn, all variables will be introduced, then the assumptions will be stated, and afterwards utility maximization will be done through the method of Lagrange multipliers and the possible intuition will be given.

The description of the variables is presented in Table 1.

Assumptions are as follows:

1. Utility function U(s, x, d, cl, c2) is continuous and twice differentiable and has a strictly quasi-concave form

2. Utility function is increasing in s as consumer would prefer larger housing unit over smaller one

3. Utility function U(s, x, d, c1, c2) is continuous and twice differentiable and has a strictly quasi-concave form

4. Utility function U(s, x, d, c1, c2) is continuous and twice differentiable and has a strictly quasi-concave form

5. Utility function is increasing in s as consumer would prefer larger housing unit over smaller one

6. Utility function is increasing in x as its consumption brings benefits to the household

7. Utility function is decreasing in d as consumers prefer to spend less time on reaching the metro station

8. Utility function is decreasing in c1 as consumers do not like noise and pollution

9. Utility function is increasing in c2 as they enjoy greater accessibility

10. Transportation cost function T(d, c1, c2) is increasing in d due to spending comparatively more energy and money on getting to metro station

11. Transportation cost function is decreasing in c1 as authorities are rational and realize the damage from negative indirect costs

12. Transportation cost function is increasing in c2 as authorities are rational and capitalize benefits brought by the station

13. Prices (price per square meter of a property) incorporate indirect costs and distances to the metro station

Constraint indicates that income can be spent rather on transport or property or compound good. Lagrange: L = U(s,x, d, cl, c2) -- A(T(d, cl, c2) + x + P(d, cl, c2) * s -- I)

F. О. C.:

1. U'_s--A*P(d,cl,c2) = 0

2. U' -- A = 0

3. U'--A*(P'*s + T;_i) = 0

4. Ud -- A* ⁽P_ci * s + Tзi) = ⁰

5. Uc2 -- A* ⁽P_C2 * S + T_(:2⁾ = ⁰

6. I = T(d, cl, c2) + x + P(d, cl, c2) * s

Price per square meter of a property should be positive. From 2) and 1) we check the rationality of next steps.

It means the positiveness of prices that is quite a rational result. This equation shows the marginal ratio a household is willing to accept to substitute consumption of square meters of property on consumption of all other goods and it is equal to the relative price.

Now we can check the impact of distance to the closest metro station on the price per square meter of a housing unit. This step will verify empirics that demonstrated negative relationship and form our prediction for the future analysis using regressions.

Using 2) and 3):

?? = _??'

By these simple calculations we observe definite negative effect of distance on price of property. Our prediction for future analysis will be based on this result - it is convenient to expect inverse relationship even without the clarifications about function forms. However, this prediction cannot influence the way we should treat the whole impact of construction of new metro stations on rental prices. As it was mentioned before, some literature indicated the fall in purchase prices and no effect for rents. Paper (Efthymiou, Antoniou, 2013) suggested possible explanations of these phenomena, and it can be captured by our model, as for the purchase price, the explanation was about negative externalities created by construction of a new line (c1 captures this impact); as for the rents, the argument was about no future benefits for tenants whose view is considered to be short term (in other words, c2 cannot affect their utility). It is reasonable to check the impact of these costs on prices.

Check for c1 by combining 2) and 4):

Check for c2 by combining 2) and 5):

Indirect costs - both positive and negative - do not give a definite result as it was for the distance to the nearest metro station. The first guess that can come to mind is the possibility of not including indirect costs c1 and c2 into transportation costs by authorities. In that case the impact will coincide with the attitude of households towards these costs (it is about the way how they are incorporated into utility). Of course, the form and the components of transportation cost function can raise doubts, but it is unlikely that gains and losses from such factors as noise and improved local accessibility are not taken into account by the market so it turns out that it is not possible to be sure about the relationship between indirect costs and property price. The second guess is related to the work from Greece that may serve a proof of an ambiguous effect that can differ from one city to another. In fact, the explanation provided by authors Efthymiou and Antoniou does not contradict the results above. If the negative impact on purchase prices is caused by negative externalities, in terms of our model there was an inverse relationship of prices P(d,c1,c2) and negative indirect costs c1was higher than (--Г_с' ₁), or the effect of marginal utility of c1 was stronger than the marginal effect from transportation costs). As for the rents, if they are not affected by possible benefits c2, it indicates that and (--Г_с'₂) were close to zero, and the explanation by Greek authors perfectly fits gotten equations. Overall, this could mean that indirect costs have high significance but it is difficult to capture them and that is why purchase prices and rents can change in any direction with respect to the strongest effect (that may be determined both by preferences of households and transportation systems policies).

To sum up, the predictions for the influence on rents state inverse relationship of distance on prices. However, indirect costs blur the total image when the impact of construction of new metro stations is tried to be seen so mathematically it can be proved that the positive correlation shown by most studies is not the only possible option and rental prices can change in any direction.

4. Methodology, data and empirical results

Research design is based on the type of data that is possible to extract from public resources. Our inputs are ten thousand strings of rental offers all over Moscow and small cities nearby from WinNER database - the information base specialized on sale and rents of real estate, on secondary market and new buildings, apartments and country houses, commercial and foreign real estate. The main aim is to include as much housing attributes and other factors as we can. However, the research faces the limitation from the data, that is why many factors were found manually from Yandex maps (as, for example, one of the most interesting factors - the proximity to the metro station). Level of the study is one of the most significant bounds as it is not possible to get the access to such technology as ArcGIS platform, a software for building GIS at any level, that helps to use geographic information to conduct analysis and simplifies data collection. After all transformations, the inputs contain names of metro stations and of streets, house numbers, rental prices in rubles, distances to the closest metro station measured in kilometers, short descriptions of each apartment, areas in square meters, numbers of storeys, floor numbers and the presence of a park or forest of any size near the house, as well as the presence of a river or lake of any origin. This database dates back to 2018, the time frame is important here, because it allows to determine the stage of construction of the metro. Flat descriptions allow to make sure that landlords are well aware of the future opening of stations. In some comments on offers, there can be seen the approximate opening date of the new station, as well as the time the road to it will take. Examining the descriptions afford to gain the confidence in the fact that the data at least partially incorporates the information on the openings within next two years, that is also confirmed by mayor's speech of that time. Thus, in the field of our research there are plenty of new metro stations: Belomorskaya, Rasskazovka, Novoperedelkino, Borovskoye Shosse, Solntsevo, Govorovo, Ozyornaya, Michurinsky Prospekt, Seligerskaya, Verkhnie Likhobory, Okruzhnaya, Delovoy Tsentr, Shelepikha, Khoroshyovskaya, CSKA, Petrovsky Park, Savyolovskaya from 2018 were in the process of opening; Filatov Lug, Prokshino, Olkhovaya, Kommunarka, Kosino, Ulitsa Dmitriyevskogo, Lukhmanovskaya, Nekrasovka were also known to be finished by 2019. The stations that we are interested in are located on the pink line, as well as on the northern part of the green and light green lines, and on the southern part of the red and yellow lines, withal on the Large circle line that duplicates some part of the yellow line (Fig.3). Housing related to these metro stations will be further assigned to a control group excluding the ones that repeat the already existing stations. Thus, in most cases, Moscow is expanding in terms of metro construction in the direction from the center.

Fig. 3 Development of Moscow metro system until 2023.

Source: Complex of urban planning policy and construction of the city of Moscow

From methods applied in modern literature, the hedonic model is chosen for the further analysis. It means that rental price is decomposed into characteristics a flat owns. Hedonic price regressions are usually estimated using secondary data on prices and housing attributes. They are based on the finding values that are formed by people's willingness to pay for a commodity when different characteristics change. Such good as a flat is considered to be heterogenous, meaning that in general, pure demand on it is too difficult to estimate that is why attributes are taken to investigate prices or elasticities in case of logarithmic model that are a common practice for real estate valuation. Advantages of hedonic price models include relative simplicity in terms of obtaining the data, cross-section or panel, and actuality, that is conditioned by actual choices of consumers. Of course, cross-sectional data that is used for this research has more limitations as it does not allow to consider the changes before/after metro development, but for this particular case it allows to capture the moment of a final stage of construction when all the dates are known.

Research is designed in the following manner. Each offer is assigned its own serial number starting from 1 and ending with 10000. Five hundred observations are taken randomly using function rand(). Resulting sample is divided into two groups - treatment and control group. Each group is tested on the best functional form, on significance of each variable and on the presence of heterogeneity, the same procedure is done with the whole sample. These actions will help to evaluate the impact of each characteristic, including the proximity to the metro station, and also, they will help to understand whether the renters have actually included the presence of a future station in the price.

Table 2 demonstrates the variables used to find the best model specifications in all further regressions

As for the hypotheses to be tested, the focus of this work is primarily on studying the difference between the rental offers for apartments that already have an access to a metro station and those who will get it in coming years. Moreover, the inverse relationship of distance on prices will be verified.

Hypothesis 1: there is a difference between the rental offers for apartments that already have an access to a metro station and those who will get it in 2018/2019 (H1: Dummy_for_metro_station is significant).

Hypothesis 2: there is an inverse relationship of distance to the closest metro station on rental prices (H2: Distance_in_km is significant and has a negative coefficient).

After the randomization, the treatment group (apartments from this group do not have a metro station nearby, but will have it in 1-2 years) consists of 126 observations. The mean value for the rental prices is 39162.70 rubles, the maximum is 300000 rubles, the minimum is 11000 rubles, while standard deviation is estimated around 29532.34 rubles.

Looking for the best model raises questions concerning possible model misspecification. Primarily, there are two reasons for the existence of this problem: rather the formula is wrong or the list of explanatory variables is wrong. To escape the first problem, all functional forms were tried including linear, semi-logarithmic and logarithmic forms. The second problem implies omitting an important explanatory variable or including unnecessary explanatory variable. Omitting a variable means that, other things being equal, estimated coefficients are biased, standard errors, t tests, and F test are invalid. This problem can be one of the most probable limitations of our analysis due to the existence of many attributes that were not taken into account. As for the second misspecification, the consequences of it are that, other things being equal, estimated coefficients are unbiased, standard errors, t tests, and F test are valid but efficiency is lower. However, it can be hard to determine, which explanatory variables are insignificant and should be excluded. Simply excluding all variables with insignificant coefficients will lead to an error because of the presence of multicollinearity. Therefore, the decision to exclude variables is made on the basis of an F-test for the joint explanatory power of several variables (F-test for linear restrictions). Joint explanatory test for dummy variables related to differentiating the last and first floors showed their insignificance for all functional forms that is why they were excluded from the analysis. However, their effects can be seen in annexes at the end of the paper.

Treatment group's best functional form was found to be:

log(rental prices) =

= a 4 Я ¦ log (distance to the closest metro station) 4

+ at * dummy for parks 4 ip * dummy for lake or river 4 i/> * floor 4

40 * number of storeys 4 у * log (appartment area)

Table 3 presents the resulting coefficients, their standard errors, t-statistics and p-values. T-tests state as a null hypothesis the insignificance of a coefficient. By these tests we reject the null hypotheses on the 5% significance level for the constant, for the logarithm of distance to the closest metro station and for the logarithm of a flat area as their p-values (0, 0.04 and 0) are lower than 5%. For a treatment group, one percent increase in distance on average leaded to 0.04% decrease in rental price while 1% increase in flat area leaded to 0.7% increase in rent. Hypothesis 2 about verification of the sign of change caused by proximity to metro is confirmed

Table 3

Treatment group

by this model. It should be noted that for all forms - linear, semilogarithmic for area and for distance separately and together, logarithmic - F-test has always shown joint significance of variables. However, in semilogarithmic and linear forms variables were insignificant separately that is the sign for multicollinearity. Multicollinearity can be caused by high correlation between explanatory variables, by small sample size, small mean standard deviations of explanatory variables and large population variance of the disturbance term. The presence of multicollinearity does not mean that the model is misspecified, Gauss-Markov theorem holds, estimates are unbiased, consistent and efficient, while statistical tests and standard errors are valid, but standard errors of estimates are higher and, with them, t-statistics are lower. Cancelling dummies for first and last floor and changing model specification seem to improve the situation.

One more problem that could arise is a problem of heteroscedasticity. For real estate market analysis, it is a common case when the scale of different economic variables changes in the same direction, they move in size together. This is also known to be a widespread situation for crosssectional data. The reason is that the variance of omitted variables and measurement errors, which together determine the values of the disturbance term, increase. Consequences of that problem are not plausible as standard errors of the regression coefficients are estimated wrongly leading to invalid t-tests and F-test. Heteroscedasticity can be tested using Breush-Pagan- Godfrey and White tests. Breush-Pagan-Godfrey test states homoscedasticity as a null hypothesis. It tests whether the variance of error terms from a regression is dependent on the values of the independent variables. After computing residuals, auxiliary regression is run and F- statistic of 0.93 is got using sample size and coefficient of determination. The p-value for the F- statistic (8,117) is 0.4945 that is much higher than significance level 0.05, thus, the null hypothesis cannot be rejected. White test is used for the detection of heteroscedasticity of the general type and also states homoscedasticity as a null hypothesis. It is very similar to the previous test but it takes into account much more cross-products. P-value for White test is 0.1681, confirming absence of heteroscedasticity. Now let's proceed to the control group.